This notebook contains the statistical analyses for the SPPT dataset.
It takes as input the following files:
For additional details we refer to the README.docx file of the repository, and the paper.
Import packages:
library(readr)
library(tidyverse)
library(tibble)
library(dplyr)
library(readxl)
library(jtools)
library(psych)
library(ggplot2)
library(rstatix)
library(huxtable)
library(caret)Load beta estimates:
# Load brain data
df_betas <- read_csv("brain/df_betas_repo.csv",col_types = cols())
# Subjects as factor
df_betas$ID <- factor(df_betas$ID)
# Get block + case from trial string
df_betas <- add_column(df_betas, block =sapply(strsplit(as.character(df_betas$case), "\\Case"), `[`, 1))
df_betas$case <- sapply(strsplit(as.character(df_betas$case), "\\Case"), `[`, 2)
# Remove case 51 because of an error with stimulus presentation for this case
remove_case <- c(51)
df_betas <- df_betas[!df_betas$case %in% remove_case, ]Center estimates per participant per VOI:
center <- function(x, na.rm = TRUE) (x - mean(x, na.rm = na.rm))
df_betas_centered <- df_betas %>%
group_by(ID) %>%
mutate_at(vars(matches('mask')), center)Add answers (case, choice, stars - confidence rating):
# Load and concatenate data
ff <- list.files("task-answers", pattern = "NNIPanswers.txt", recursive = TRUE, full.names = TRUE)
star_data <- do.call(rbind, lapply(ff, read.delim))
# Add subject number
sub_nr <- rep(2:37, each=45)
sub_nr <- sprintf("%02d", sub_nr)
ID <- paste("sub-",sub_nr, sep="")
star_data <- cbind(ID,star_data)
# Rename and fix columns
names(star_data)[names(star_data) == "Case"] <- "case"
star_data <- star_data[c("ID","case","Stars")]
star_data <- star_data %>% mutate(case = as.character(case))
# Combine data frames
all_data_df <- star_data %>% right_join(df_betas_centered, by=c("ID","case"))
all_data_df <- all_data_df %>% relocate(Stars, .after=last_col())Add stock metrics:
# Add case info
Stock_metrics <- read_excel("Stock_metrics.xlsx")
# Convert case to factor
Stock_metrics$case <- factor(Stock_metrics$case)
# Match by case
all_data_df <- cbind(all_data_df, Stock_metrics[match(all_data_df$case, Stock_metrics$case),])
all_data_df <- all_data_df[, !duplicated(colnames(all_data_df))]
# Log-transform CMC because it is highly skewed
all_data_df$log_CMC <- log(all_data_df$CMC)
# Add correct
all_data_df$cor_p1Y <- ifelse(all_data_df$choice==all_data_df$Rel_p1Y_pos, 1, 0) #Correct prediction at 1-year in the future (1 if equal, 0 if not)And Survey scores: (This also removes subjects 17,18 - these subjects are removed because of excessive head-motion)
# Import data
survey_scores <- read_csv("Survey_scores.csv",col_types = cols())
# Join to all_data
all_data_df <- merge(all_data_df,survey_scores, by="ID")Make names unique
names(all_data_df)<-make.names(names(all_data_df),unique = TRUE)Calculate average per participant
Sub_mean <- all_data_df %>%
group_by(ID) %>%
summarise_at(vars(-case,-block,-trial), list(~ mean(., na.rm=TRUE)))
#Sub_meanCalculate age and years of experience:
x=Sub_mean[50:53]
sapply(x, function(x) c( "Stand dev" = sd(x),
"Mean"= mean(x,na.rm=TRUE),
"n" = length(x),
"Median" = median(x),
"CoeffofVariation" = sd(x)/mean(x,na.rm=TRUE),
"Minimum" = min(x),
"Maximun" = max(x),
"Upper Quantile" = quantile(x,1),
"LowerQuartile" = quantile(x,0)
)
) Age finance_exp asset_exp equity_exp
Stand dev 9.0894250 9.9644757 9.3608702 9.552195
Mean 47.5588235 19.2205882 12.3676471 14.926471
n 34.0000000 34.0000000 34.0000000 34.000000
Median 49.0000000 20.0000000 10.0000000 13.500000
CoeffofVariation 0.1911196 0.5184272 0.7568837 0.639950
Minimum 29.0000000 0.0000000 0.0000000 0.000000
Maximun 66.0000000 40.0000000 40.0000000 40.000000
Upper Quantile.100% 66.0000000 40.0000000 40.0000000 40.000000
LowerQuartile.0% 29.0000000 0.0000000 0.0000000 0.000000Calculate self-reported difficulty/ realism and test if above mid-point of scale:
x=Sub_mean[55:56]
sapply(x, function(x) c( "Stand dev" = sd(x),
"Mean"= mean(x,na.rm=TRUE),
"n" = length(x),
"Median" = median(x),
"CoeffofVariation" = sd(x)/mean(x,na.rm=TRUE),
"Minimum" = min(x),
"Maximun" = max(x),
"Upper Quantile" = quantile(x,1),
"LowerQuartile" = quantile(x,0)
)
) realistic difficult
Stand dev 1.4129526 0.9883456
Mean 4.0588235 4.4117647
n 34.0000000 34.0000000
Median 4.0000000 5.0000000
CoeffofVariation 0.3481187 0.2240250
Minimum 1.0000000 2.0000000
Maximun 7.0000000 6.0000000
Upper Quantile.100% 7.0000000 6.0000000
LowerQuartile.0% 1.0000000 2.0000000t.test(Sub_mean$realistic, mu=4, alternative="two.sided")
One Sample t-test
data: Sub_mean$realistic
t = 0.24275, df = 33, p-value = 0.8097
alternative hypothesis: true mean is not equal to 4
95 percent confidence interval:
3.565821 4.551826
sample estimates:
mean of x
4.058824 t.test(Sub_mean$difficult, mu=4, alternative="two.sided")
One Sample t-test
data: Sub_mean$difficult
t = 2.4293, df = 33, p-value = 0.02074
alternative hypothesis: true mean is not equal to 4
95 percent confidence interval:
4.066915 4.756615
sample estimates:
mean of x
4.411765 Self-reported importance of information screens:
x=Sub_mean[61:65]
sapply(x, function(x) c( "Stand dev" = sd(x),
"Mean"= mean(x,na.rm=TRUE),
"n" = length(x),
"Median" = median(x),
"CoeffofVariation" = sd(x)/mean(x,na.rm=TRUE),
"Minimum" = min(x),
"Maximun" = max(x),
"Upper Quantile" = quantile(x,1),
"LowerQuartile" = quantile(x,0)
)
) prof_imp fund_imp rela_imp graph_imp news_imp
Stand dev 1.225109 1.0294245 1.1930428 0.9650764 1.2878840
Mean 1.882353 3.9705882 2.9705882 2.9117647 3.9117647
n 34.000000 34.0000000 34.0000000 34.0000000 34.0000000
Median 1.000000 4.0000000 3.0000000 3.0000000 4.0000000
CoeffofVariation 0.650839 0.2592625 0.4016184 0.3314404 0.3292335
Minimum 1.000000 2.0000000 1.0000000 1.0000000 1.0000000
Maximun 5.000000 5.0000000 5.0000000 5.0000000 5.0000000
Upper Quantile.100% 5.000000 5.0000000 5.0000000 5.0000000 5.0000000
LowerQuartile.0% 1.000000 2.0000000 1.0000000 1.0000000 1.0000000Reshape data and conduct pairwise tests:
x=Sub_mean[c(1,61:65)]
importance_scores <- pivot_longer(x, cols=-ID,names_to='Screen',values_to='Importance')
pwc <- importance_scores %>%
pairwise_t_test(
Importance ~ Screen, paired = TRUE,
p.adjust.method = "bonferroni"
)
pwcCalculate average per case
Case_mean <- all_data_df %>%
group_by(case) %>%
summarise_at(vars(-ID,-block,-trial), funs(mean(., na.rm=TRUE)))
Case_meanAverage prediction and correct prediction
x=Case_mean[c(32,49)]
sapply(x, function(x) c( "Stand dev" = sd(x),
"Mean"= mean(x,na.rm=TRUE),
"n" = length(x),
"Median" = median(x),
"CoeffofVariation" = sd(x)/mean(x,na.rm=TRUE),
"Minimum" = min(x),
"Maximun" = max(x),
"Upper Quantile" = quantile(x,1),
"LowerQuartile" = quantile(x,0)
)
) choice cor_p1Y
Stand dev 0.24337304 0.24196615
Mean 0.49547657 0.52622524
n 44.00000000 44.00000000
Median 0.52941176 0.56684492
CoeffofVariation 0.49118981 0.45981479
Minimum 0.05882353 0.05882353
Maximun 0.88235294 0.88235294
Upper Quantile.100% 0.88235294 0.88235294
LowerQuartile.0% 0.05882353 0.05882353Logistic regression: predictions to future market performance
Rel_p1Y_pos = Relative performance 1 year in the future (underperforming/overperforming, coded as 0/1)
formula=paste("Rel_p1Y_pos~choice",sep="")
print(formula)[1] "Rel_p1Y_pos~choice"out_choice <- glm(formula,data=Case_mean,family = binomial(link = logit))
print(summ(out_choice,digits = getOption("jtools-digits", default = 3), scale=TRUE))MODEL INFO:
Observations: 44
Dependent Variable: Rel_p1Y_pos
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(1) = 0.525, p = 0.469
Pseudo-R² (Cragg-Uhler) = 0.016
Pseudo-R² (McFadden) = 0.009
AIC = 64.472, BIC = 68.041
Standard errors: MLE
---------------------------------------------------
Est. S.E. z val. p
----------------- -------- ------- -------- -------
(Intercept) -0.000 0.303 -0.000 1.000
choice 0.222 0.308 0.720 0.471
---------------------------------------------------
Continuous predictors are mean-centered and scaled by 1 s.d. The outcome variable remains in its original units.Correlation between self-reported confidence (Stars) and prediction accuracy
cor.test(Case_mean$Stars, Case_mean$cor_p1Y)
Pearson's product-moment correlation
data: Case_mean$Stars and Case_mean$cor_p1Y
t = 0.15821, df = 42, p-value = 0.875
alternative hypothesis: true correlation is not equal to 0
95 percent confidence interval:
-0.2744637 0.3189745
sample estimates:
cor
0.02440524 Select columns:
nacc <- Case_mean[, c(4,9,14,19,24)]
KMO(r=cor(nacc))Kaiser-Meyer-Olkin factor adequacy
Call: KMO(r = cor(nacc))
Overall MSA = 0.74
MSA for each item =
fund.Bilat_NAcc_mask graph.Bilat_NAcc_mask news.Bilat_NAcc_mask prof.Bilat_NAcc_mask rela.Bilat_NAcc_mask
0.73 0.80 0.70 0.69 0.77 Positive correlation accross information screens:
median(cor(nacc))[1] 0.5581268Factor analysis:
fit <- factanal(nacc, 2,rotation = "promax", scores='regression')
print(fit, digits=3, cutoff=0.3, sort=TRUE)
Call:
factanal(x = nacc, factors = 2, scores = "regression", rotation = "promax")
Uniquenesses:
fund.Bilat_NAcc_mask graph.Bilat_NAcc_mask news.Bilat_NAcc_mask prof.Bilat_NAcc_mask rela.Bilat_NAcc_mask
0.172 0.371 0.396 0.388 0.386
Loadings:
Factor1 Factor2
fund.Bilat_NAcc_mask 0.668 0.381
graph.Bilat_NAcc_mask 0.675
prof.Bilat_NAcc_mask 0.865
news.Bilat_NAcc_mask 0.831
rela.Bilat_NAcc_mask 0.702
Factor1 Factor2
SS loadings 1.692 1.432
Proportion Var 0.338 0.286
Cumulative Var 0.338 0.625
Factor Correlations:
Factor1 Factor2
Factor1 1.000 0.463
Factor2 0.463 1.000
Test of the hypothesis that 2 factors are sufficient.
The chi square statistic is 0.36 on 1 degree of freedom.
The p-value is 0.549 Add factor scores, creating df for regression:
NAcc_factor_scores <- data.frame(fit$scores)
reg_df <- merge(Case_mean,NAcc_factor_scores,by="row.names",all.x=TRUE)Regression analysis per information screen (for Table 2):
formula=paste("Rel_p1Y_pos~log_CMC+prof.Bilat_NAcc_mask+prof.Bilat_MPFC_mask+prof.Bilat_AIns_mask",sep="")
print(formula)[1] "Rel_p1Y_pos~log_CMC+prof.Bilat_NAcc_mask+prof.Bilat_MPFC_mask+prof.Bilat_AIns_mask"out_prof <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_prof,digits = getOption("jtools-digits", default = 2),scale=TRUE))MODEL INFO:
Observations: 44
Dependent Variable: Rel_p1Y_pos
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(4) = 10.54, p = 0.03
Pseudo-R² (Cragg-Uhler) = 0.28
Pseudo-R² (McFadden) = 0.17
AIC = 60.45, BIC = 69.37
Standard errors: MLE
---------------------------------------------------------
Est. S.E. z val. p
-------------------------- ------- ------ -------- ------
(Intercept) 0.01 0.34 0.03 0.98
log_CMC 0.31 0.35 0.88 0.38
prof.Bilat_NAcc_mask 1.36 0.62 2.20 0.03
prof.Bilat_MPFC_mask 0.28 0.61 0.47 0.64
prof.Bilat_AIns_mask -0.75 0.55 -1.37 0.17
---------------------------------------------------------
Continuous predictors are mean-centered and scaled by 1 s.d. The outcome variable remains in its original units.formula=paste("Rel_p1Y_pos~Rel_m3Y_pos+graph.Bilat_NAcc_mask+graph.Bilat_MPFC_mask+graph.Bilat_AIns_mask",sep="")
print(formula)[1] "Rel_p1Y_pos~Rel_m3Y_pos+graph.Bilat_NAcc_mask+graph.Bilat_MPFC_mask+graph.Bilat_AIns_mask"out_graph <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_graph,digits = getOption("jtools-digits", default = 2),scale=TRUE))MODEL INFO:
Observations: 44
Dependent Variable: Rel_p1Y_pos
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(4) = 12.72, p = 0.01
Pseudo-R² (Cragg-Uhler) = 0.33
Pseudo-R² (McFadden) = 0.21
AIC = 58.28, BIC = 67.20
Standard errors: MLE
----------------------------------------------------------
Est. S.E. z val. p
--------------------------- ------- ------ -------- ------
(Intercept) -0.70 0.64 -1.09 0.28
Rel_m3Y_pos 1.06 0.85 1.25 0.21
graph.Bilat_NAcc_mask 2.14 0.86 2.47 0.01
graph.Bilat_MPFC_mask -0.83 0.59 -1.41 0.16
graph.Bilat_AIns_mask -0.30 0.75 -0.39 0.69
----------------------------------------------------------
Continuous predictors are mean-centered and scaled by 1 s.d. The outcome variable remains in its original units.formula=paste("Rel_p1Y_pos~f_s+f_ebit+f_ebitmargin+f_roe+f_roic+fund.Bilat_NAcc_mask+fund.Bilat_MPFC_mask+fund.Bilat_AIns_mask",sep="")
print(formula)[1] "Rel_p1Y_pos~f_s+f_ebit+f_ebitmargin+f_roe+f_roic+fund.Bilat_NAcc_mask+fund.Bilat_MPFC_mask+fund.Bilat_AIns_mask"out_fund <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_fund,digits = getOption("jtools-digits", default = 2),scale=TRUE))MODEL INFO:
Observations: 44
Dependent Variable: Rel_p1Y_pos
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(8) = 6.98, p = 0.54
Pseudo-R² (Cragg-Uhler) = 0.20
Pseudo-R² (McFadden) = 0.11
AIC = 72.01, BIC = 88.07
Standard errors: MLE
---------------------------------------------------------
Est. S.E. z val. p
-------------------------- ------- ------ -------- ------
(Intercept) -0.00 0.33 -0.01 0.99
f_s -0.29 0.47 -0.62 0.54
f_ebit 0.06 0.43 0.13 0.89
f_ebitmargin 0.26 0.49 0.52 0.60
f_roe -0.29 0.73 -0.40 0.69
f_roic -0.15 0.59 -0.25 0.80
fund.Bilat_NAcc_mask 1.46 0.68 2.13 0.03
fund.Bilat_MPFC_mask 0.11 0.51 0.21 0.83
fund.Bilat_AIns_mask -1.23 0.77 -1.61 0.11
---------------------------------------------------------
Continuous predictors are mean-centered and scaled by 1 s.d. The outcome variable remains in its original units.formula=paste("Rel_p1Y_pos~rv_evs+rv_ebitda+rv_pe+rv_pb+rela.Bilat_NAcc_mask+rela.Bilat_MPFC_mask+rela.Bilat_AIns_mask",sep="")
print(formula)[1] "Rel_p1Y_pos~rv_evs+rv_ebitda+rv_pe+rv_pb+rela.Bilat_NAcc_mask+rela.Bilat_MPFC_mask+rela.Bilat_AIns_mask"out_rela <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_rela,digits = getOption("jtools-digits", default = 2),scale=TRUE))MODEL INFO:
Observations: 44
Dependent Variable: Rel_p1Y_pos
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(7) = 8.35, p = 0.30
Pseudo-R² (Cragg-Uhler) = 0.23
Pseudo-R² (McFadden) = 0.14
AIC = 68.65, BIC = 82.92
Standard errors: MLE
---------------------------------------------------------
Est. S.E. z val. p
-------------------------- ------- ------ -------- ------
(Intercept) 0.08 0.45 0.19 0.85
rv_evs 0.27 0.65 0.41 0.68
rv_ebitda 1.05 0.70 1.51 0.13
rv_pe -0.81 0.70 -1.16 0.25
rv_pb 0.78 1.96 0.40 0.69
rela.Bilat_NAcc_mask 0.65 0.57 1.14 0.26
rela.Bilat_MPFC_mask 0.42 0.50 0.85 0.40
rela.Bilat_AIns_mask -0.86 0.58 -1.47 0.14
---------------------------------------------------------
Continuous predictors are mean-centered and scaled by 1 s.d. The outcome variable remains in its original units.formula=paste("Rel_p1Y_pos~news_incgr+news.Bilat_NAcc_mask+news.Bilat_MPFC_mask+news.Bilat_AIns_mask",sep="")
print(formula)[1] "Rel_p1Y_pos~news_incgr+news.Bilat_NAcc_mask+news.Bilat_MPFC_mask+news.Bilat_AIns_mask"out_news <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_news,digits = getOption("jtools-digits", default = 2),scale=TRUE))MODEL INFO:
Observations: 44
Dependent Variable: Rel_p1Y_pos
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(4) = 4.10, p = 0.39
Pseudo-R² (Cragg-Uhler) = 0.12
Pseudo-R² (McFadden) = 0.07
AIC = 66.90, BIC = 75.82
Standard errors: MLE
---------------------------------------------------------
Est. S.E. z val. p
-------------------------- ------- ------ -------- ------
(Intercept) 0.14 0.39 0.37 0.71
news_incgr 1.16 1.48 0.78 0.43
news.Bilat_NAcc_mask -0.34 0.49 -0.71 0.48
news.Bilat_MPFC_mask 0.24 0.49 0.49 0.62
news.Bilat_AIns_mask 0.41 0.50 0.82 0.41
---------------------------------------------------------
Continuous predictors are mean-centered and scaled by 1 s.d. The outcome variable remains in its original units.Output table:
coef_names <- c("Current Market Capitalization" = "log_CMC", "Profile - Nacc" = "prof.Bilat_NAcc_mask", "Profile - MPFC" = "prof.Bilat_MPFC_mask", "Profile - AIns" = "prof.Bilat_AIns_mask",
"36M Performance"='Rel_m3Y_pos', "Graph - Nacc" = "graph.Bilat_NAcc_mask", "Graph - MPFC" = "graph.Bilat_MPFC_mask", "Graph - AIns" = "graph.Bilat_AIns_mask",
"Sales" ="f_s", "EBIT" = "f_ebit", "EBIT Margin (%)" = "f_ebitmargin","ROE (%)" = "f_roe","ROIC (%)" = "f_roic",
"Fundamentals - Nacc" = "fund.Bilat_NAcc_mask", "Fundamentals - MPFC" = "fund.Bilat_MPFC_mask", "Fundamentals - AIns" = "fund.Bilat_AIns_mask",
"EV/Sales" = "rv_evs","EV/EBITDA" = "rv_ebitda","P/E" = "rv_pe","P/B" = "rv_pb",
"Relative Valuation - Nacc" = "rela.Bilat_NAcc_mask", "Relative Valuation - MPFC" = "rela.Bilat_MPFC_mask", "Relative Valuation - AIns" = "rela.Bilat_AIns_mask",
"Income growth" = "news_incgr","News item - Nacc" = "news.Bilat_NAcc_mask", "News item - MPFC" = "news.Bilat_MPFC_mask", "News item - AIns" = "news.Bilat_AIns_mask" )
# Uncomment to create table
# quick_docx(export_summs(out_prof,out_graph,out_fund,out_rela,out_news, scale = TRUE,coefs = coef_names,error_pos = c("same"),
# model.names=c("Company Profile","Price Graph","Fundamentals","Relative Valuation","News Item"),
# statistics= c(N = "nobs", AIC = "AIC", "Pseudo R2 (McFadden)" = "pseudo.r.squared.mcfadden")), file='Table-2.docx')Only neural predictors:
formula=paste("Rel_p1Y_pos~prof.Bilat_NAcc_mask+prof.Bilat_MPFC_mask+prof.Bilat_AIns_mask",sep="")
print(formula)[1] "Rel_p1Y_pos~prof.Bilat_NAcc_mask+prof.Bilat_MPFC_mask+prof.Bilat_AIns_mask"out_prof_neural <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_prof_neural,digits = getOption("jtools-digits", default = 2),scale=TRUE))MODEL INFO:
Observations: 44
Dependent Variable: Rel_p1Y_pos
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(3) = 9.76, p = 0.02
Pseudo-R² (Cragg-Uhler) = 0.27
Pseudo-R² (McFadden) = 0.16
AIC = 59.24, BIC = 66.38
Standard errors: MLE
---------------------------------------------------------
Est. S.E. z val. p
-------------------------- ------- ------ -------- ------
(Intercept) 0.01 0.34 0.02 0.98
prof.Bilat_NAcc_mask 1.37 0.61 2.23 0.03
prof.Bilat_MPFC_mask 0.16 0.59 0.27 0.79
prof.Bilat_AIns_mask -0.68 0.54 -1.26 0.21
---------------------------------------------------------
Continuous predictors are mean-centered and scaled by 1 s.d. The outcome variable remains in its original units.formula=paste("Rel_p1Y_pos~graph.Bilat_NAcc_mask+graph.Bilat_MPFC_mask+graph.Bilat_AIns_mask",sep="")
print(formula)[1] "Rel_p1Y_pos~graph.Bilat_NAcc_mask+graph.Bilat_MPFC_mask+graph.Bilat_AIns_mask"out_graph_neural <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_graph_neural,digits = getOption("jtools-digits", default = 2),scale=TRUE))MODEL INFO:
Observations: 44
Dependent Variable: Rel_p1Y_pos
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(3) = 11.06, p = 0.01
Pseudo-R² (Cragg-Uhler) = 0.30
Pseudo-R² (McFadden) = 0.18
AIC = 57.94, BIC = 65.07
Standard errors: MLE
----------------------------------------------------------
Est. S.E. z val. p
--------------------------- ------- ------ -------- ------
(Intercept) -0.03 0.34 -0.10 0.92
graph.Bilat_NAcc_mask 1.80 0.77 2.32 0.02
graph.Bilat_MPFC_mask -0.89 0.58 -1.53 0.13
graph.Bilat_AIns_mask -0.19 0.72 -0.26 0.80
----------------------------------------------------------
Continuous predictors are mean-centered and scaled by 1 s.d. The outcome variable remains in its original units.formula=paste("Rel_p1Y_pos~fund.Bilat_NAcc_mask+fund.Bilat_MPFC_mask+fund.Bilat_AIns_mask",sep="")
print(formula)[1] "Rel_p1Y_pos~fund.Bilat_NAcc_mask+fund.Bilat_MPFC_mask+fund.Bilat_AIns_mask"out_fund_neural <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_fund_neural,digits = getOption("jtools-digits", default = 2),scale=TRUE))MODEL INFO:
Observations: 44
Dependent Variable: Rel_p1Y_pos
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(3) = 5.55, p = 0.14
Pseudo-R² (Cragg-Uhler) = 0.16
Pseudo-R² (McFadden) = 0.09
AIC = 63.45, BIC = 70.59
Standard errors: MLE
---------------------------------------------------------
Est. S.E. z val. p
-------------------------- ------- ------ -------- ------
(Intercept) -0.01 0.32 -0.03 0.98
fund.Bilat_NAcc_mask 1.38 0.65 2.12 0.03
fund.Bilat_MPFC_mask 0.17 0.49 0.34 0.73
fund.Bilat_AIns_mask -1.11 0.73 -1.52 0.13
---------------------------------------------------------
Continuous predictors are mean-centered and scaled by 1 s.d. The outcome variable remains in its original units.formula=paste("Rel_p1Y_pos~rela.Bilat_NAcc_mask+rela.Bilat_MPFC_mask+rela.Bilat_AIns_mask",sep="")
print(formula)[1] "Rel_p1Y_pos~rela.Bilat_NAcc_mask+rela.Bilat_MPFC_mask+rela.Bilat_AIns_mask"out_rela_neural <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_rela_neural,digits = getOption("jtools-digits", default = 2),scale=TRUE))MODEL INFO:
Observations: 44
Dependent Variable: Rel_p1Y_pos
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(3) = 3.93, p = 0.27
Pseudo-R² (Cragg-Uhler) = 0.11
Pseudo-R² (McFadden) = 0.06
AIC = 65.06, BIC = 72.20
Standard errors: MLE
---------------------------------------------------------
Est. S.E. z val. p
-------------------------- ------- ------ -------- ------
(Intercept) -0.00 0.32 -0.02 0.99
rela.Bilat_NAcc_mask 0.79 0.56 1.40 0.16
rela.Bilat_MPFC_mask 0.37 0.48 0.76 0.45
rela.Bilat_AIns_mask -0.87 0.57 -1.53 0.12
---------------------------------------------------------
Continuous predictors are mean-centered and scaled by 1 s.d. The outcome variable remains in its original units.formula=paste("Rel_p1Y_pos~news.Bilat_NAcc_mask+news.Bilat_MPFC_mask+news.Bilat_AIns_mask",sep="")
print(formula)[1] "Rel_p1Y_pos~news.Bilat_NAcc_mask+news.Bilat_MPFC_mask+news.Bilat_AIns_mask"out_news_neural <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_news_neural,digits = getOption("jtools-digits", default = 2),scale=TRUE))MODEL INFO:
Observations: 44
Dependent Variable: Rel_p1Y_pos
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(3) = 2.56, p = 0.46
Pseudo-R² (Cragg-Uhler) = 0.08
Pseudo-R² (McFadden) = 0.04
AIC = 66.43, BIC = 73.57
Standard errors: MLE
---------------------------------------------------------
Est. S.E. z val. p
-------------------------- ------- ------ -------- ------
(Intercept) -0.00 0.31 -0.01 0.99
news.Bilat_NAcc_mask -0.41 0.48 -0.85 0.39
news.Bilat_MPFC_mask 0.31 0.46 0.67 0.50
news.Bilat_AIns_mask 0.51 0.47 1.10 0.27
---------------------------------------------------------
Continuous predictors are mean-centered and scaled by 1 s.d. The outcome variable remains in its original units.Reduced models with only NAcc predictors (without MPFC and AIns) and stock metrics for each screen:
formula=paste("Rel_p1Y_pos~log_CMC+prof.Bilat_NAcc_mask",sep="")
print(formula)[1] "Rel_p1Y_pos~log_CMC+prof.Bilat_NAcc_mask"out_prof_NAcc <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_prof_NAcc,digits = getOption("jtools-digits", default = 2),scale=TRUE))MODEL INFO:
Observations: 44
Dependent Variable: Rel_p1Y_pos
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(2) = 8.57, p = 0.01
Pseudo-R² (Cragg-Uhler) = 0.24
Pseudo-R² (McFadden) = 0.14
AIC = 58.42, BIC = 63.78
Standard errors: MLE
--------------------------------------------------------
Est. S.E. z val. p
-------------------------- ------ ------ -------- ------
(Intercept) 0.02 0.33 0.05 0.96
log_CMC 0.26 0.33 0.77 0.44
prof.Bilat_NAcc_mask 0.98 0.38 2.59 0.01
--------------------------------------------------------
Continuous predictors are mean-centered and scaled by 1 s.d. The outcome variable remains in its original units.formula=paste("Rel_p1Y_pos~Rel_m3Y_pos+graph.Bilat_NAcc_mask",sep="")
print(formula)[1] "Rel_p1Y_pos~Rel_m3Y_pos+graph.Bilat_NAcc_mask"out_graph_NAcc <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_graph_NAcc,digits = getOption("jtools-digits", default = 2),scale=TRUE))MODEL INFO:
Observations: 44
Dependent Variable: Rel_p1Y_pos
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(2) = 9.32, p = 0.01
Pseudo-R² (Cragg-Uhler) = 0.25
Pseudo-R² (McFadden) = 0.15
AIC = 57.68, BIC = 63.03
Standard errors: MLE
----------------------------------------------------------
Est. S.E. z val. p
--------------------------- ------- ------ -------- ------
(Intercept) -0.63 0.59 -1.06 0.29
Rel_m3Y_pos 1.08 0.80 1.34 0.18
graph.Bilat_NAcc_mask 1.23 0.47 2.59 0.01
----------------------------------------------------------
Continuous predictors are mean-centered and scaled by 1 s.d. The outcome variable remains in its original units.formula=paste("Rel_p1Y_pos~f_s+f_ebit+f_ebitmargin+f_roe+f_roic+fund.Bilat_NAcc_mask",sep="")
print(formula)[1] "Rel_p1Y_pos~f_s+f_ebit+f_ebitmargin+f_roe+f_roic+fund.Bilat_NAcc_mask"out_fund_NAcc <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_fund_NAcc,digits = getOption("jtools-digits", default = 2),scale=TRUE))MODEL INFO:
Observations: 44
Dependent Variable: Rel_p1Y_pos
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(6) = 3.72, p = 0.71
Pseudo-R² (Cragg-Uhler) = 0.11
Pseudo-R² (McFadden) = 0.06
AIC = 71.27, BIC = 83.76
Standard errors: MLE
---------------------------------------------------------
Est. S.E. z val. p
-------------------------- ------- ------ -------- ------
(Intercept) 0.00 0.32 0.01 0.99
f_s -0.11 0.45 -0.24 0.81
f_ebit -0.09 0.41 -0.22 0.82
f_ebitmargin 0.31 0.47 0.66 0.51
f_roe -0.13 0.69 -0.19 0.85
f_roic -0.24 0.63 -0.38 0.71
fund.Bilat_NAcc_mask 0.50 0.36 1.40 0.16
---------------------------------------------------------
Continuous predictors are mean-centered and scaled by 1 s.d. The outcome variable remains in its original units.formula=paste("Rel_p1Y_pos~rv_evs+rv_ebitda+rv_pe+rv_pb+rela.Bilat_NAcc_mask",sep="")
print(formula)[1] "Rel_p1Y_pos~rv_evs+rv_ebitda+rv_pe+rv_pb+rela.Bilat_NAcc_mask"out_rela_NAcc <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_rela_NAcc,digits = getOption("jtools-digits", default = 2),scale=TRUE))MODEL INFO:
Observations: 44
Dependent Variable: Rel_p1Y_pos
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(5) = 5.87, p = 0.32
Pseudo-R² (Cragg-Uhler) = 0.17
Pseudo-R² (McFadden) = 0.10
AIC = 67.13, BIC = 77.84
Standard errors: MLE
---------------------------------------------------------
Est. S.E. z val. p
-------------------------- ------- ------ -------- ------
(Intercept) 0.11 0.49 0.23 0.82
rv_evs 0.16 0.68 0.23 0.82
rv_ebitda 0.99 0.65 1.52 0.13
rv_pe -0.81 0.64 -1.25 0.21
rv_pb 0.93 2.38 0.39 0.70
rela.Bilat_NAcc_mask 0.27 0.35 0.78 0.43
---------------------------------------------------------
Continuous predictors are mean-centered and scaled by 1 s.d. The outcome variable remains in its original units.formula=paste("Rel_p1Y_pos~news_incgr+news.Bilat_NAcc_mask",sep="")
print(formula)[1] "Rel_p1Y_pos~news_incgr+news.Bilat_NAcc_mask"out_news_NAcc <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_news_NAcc,digits = getOption("jtools-digits", default = 2),scale=TRUE))MODEL INFO:
Observations: 44
Dependent Variable: Rel_p1Y_pos
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(2) = 2.77, p = 0.25
Pseudo-R² (Cragg-Uhler) = 0.08
Pseudo-R² (McFadden) = 0.05
AIC = 64.23, BIC = 69.58
Standard errors: MLE
--------------------------------------------------------
Est. S.E. z val. p
-------------------------- ------ ------ -------- ------
(Intercept) 0.17 0.38 0.45 0.65
news_incgr 1.45 1.40 1.03 0.30
news.Bilat_NAcc_mask 0.06 0.32 0.19 0.85
--------------------------------------------------------
Continuous predictors are mean-centered and scaled by 1 s.d. The outcome variable remains in its original units.Reduced models with only NAcc predictors (without MPFC and AIns) and without stock metrics for each screen:
formula=paste("Rel_p1Y_pos~prof.Bilat_NAcc_mask",sep="")
print(formula)[1] "Rel_p1Y_pos~prof.Bilat_NAcc_mask"out_prof_NAcc2 <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_prof_NAcc2,digits = getOption("jtools-digits", default = 2),scale=TRUE))MODEL INFO:
Observations: 44
Dependent Variable: Rel_p1Y_pos
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(1) = 7.98, p = 0.00
Pseudo-R² (Cragg-Uhler) = 0.22
Pseudo-R² (McFadden) = 0.13
AIC = 57.02, BIC = 60.59
Standard errors: MLE
--------------------------------------------------------
Est. S.E. z val. p
-------------------------- ------ ------ -------- ------
(Intercept) 0.01 0.33 0.02 0.99
prof.Bilat_NAcc_mask 0.95 0.37 2.56 0.01
--------------------------------------------------------
Continuous predictors are mean-centered and scaled by 1 s.d. The outcome variable remains in its original units.formula=paste("Rel_p1Y_pos~graph.Bilat_NAcc_mask",sep="")
print(formula)[1] "Rel_p1Y_pos~graph.Bilat_NAcc_mask"out_graph_NAcc2 <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_graph_NAcc2,digits = getOption("jtools-digits", default = 2),scale=TRUE))MODEL INFO:
Observations: 44
Dependent Variable: Rel_p1Y_pos
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(1) = 7.37, p = 0.01
Pseudo-R² (Cragg-Uhler) = 0.21
Pseudo-R² (McFadden) = 0.12
AIC = 57.62, BIC = 61.19
Standard errors: MLE
---------------------------------------------------------
Est. S.E. z val. p
--------------------------- ------ ------ -------- ------
(Intercept) 0.02 0.33 0.06 0.96
graph.Bilat_NAcc_mask 0.95 0.39 2.40 0.02
---------------------------------------------------------
Continuous predictors are mean-centered and scaled by 1 s.d. The outcome variable remains in its original units.formula=paste("Rel_p1Y_pos~fund.Bilat_NAcc_mask",sep="")
print(formula)[1] "Rel_p1Y_pos~fund.Bilat_NAcc_mask"out_fund_NAcc2 <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_fund_NAcc2,digits = getOption("jtools-digits", default = 2),scale=TRUE))MODEL INFO:
Observations: 44
Dependent Variable: Rel_p1Y_pos
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(1) = 2.70, p = 0.10
Pseudo-R² (Cragg-Uhler) = 0.08
Pseudo-R² (McFadden) = 0.04
AIC = 62.30, BIC = 65.87
Standard errors: MLE
---------------------------------------------------------
Est. S.E. z val. p
-------------------------- ------- ------ -------- ------
(Intercept) -0.00 0.31 -0.01 0.99
fund.Bilat_NAcc_mask 0.53 0.34 1.55 0.12
---------------------------------------------------------
Continuous predictors are mean-centered and scaled by 1 s.d. The outcome variable remains in its original units.formula=paste("Rel_p1Y_pos~rela.Bilat_NAcc_mask",sep="")
print(formula)[1] "Rel_p1Y_pos~rela.Bilat_NAcc_mask"out_rela_NAcc2 <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_rela_NAcc2,digits = getOption("jtools-digits", default = 2),scale=TRUE))MODEL INFO:
Observations: 44
Dependent Variable: Rel_p1Y_pos
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(1) = 1.31, p = 0.25
Pseudo-R² (Cragg-Uhler) = 0.04
Pseudo-R² (McFadden) = 0.02
AIC = 63.69, BIC = 67.25
Standard errors: MLE
--------------------------------------------------------
Est. S.E. z val. p
-------------------------- ------ ------ -------- ------
(Intercept) 0.00 0.31 0.01 0.99
rela.Bilat_NAcc_mask 0.36 0.33 1.10 0.27
--------------------------------------------------------
Continuous predictors are mean-centered and scaled by 1 s.d. The outcome variable remains in its original units.formula=paste("Rel_p1Y_pos~news.Bilat_NAcc_mask",sep="")
print(formula)[1] "Rel_p1Y_pos~news.Bilat_NAcc_mask"out_news_NAcc2 <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_news_NAcc2,digits = getOption("jtools-digits", default = 2),scale=TRUE))MODEL INFO:
Observations: 44
Dependent Variable: Rel_p1Y_pos
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(1) = 0.17, p = 0.68
Pseudo-R² (Cragg-Uhler) = 0.01
Pseudo-R² (McFadden) = 0.00
AIC = 64.82, BIC = 68.39
Standard errors: MLE
---------------------------------------------------------
Est. S.E. z val. p
-------------------------- ------- ------ -------- ------
(Intercept) -0.00 0.30 -0.00 1.00
news.Bilat_NAcc_mask 0.13 0.31 0.42 0.68
---------------------------------------------------------
Continuous predictors are mean-centered and scaled by 1 s.d. The outcome variable remains in its original units.T-tests for NAcc activity (underperforming/overperforming)
t.test(prof.Bilat_NAcc_mask~Rel_p1Y_pos, data=reg_df,var.equal=TRUE)
Two Sample t-test
data: prof.Bilat_NAcc_mask by Rel_p1Y_pos
t = -2.9335, df = 42, p-value = 0.005409
alternative hypothesis: true difference in means between group 0 and group 1 is not equal to 0
95 percent confidence interval:
-0.12753448 -0.02357879
sample estimates:
mean in group 0 mean in group 1
-0.03770480 0.03785183 t.test(graph.Bilat_NAcc_mask~Rel_p1Y_pos, data=reg_df,var.equal=TRUE)
Two Sample t-test
data: graph.Bilat_NAcc_mask by Rel_p1Y_pos
t = -2.7664, df = 42, p-value = 0.00839
alternative hypothesis: true difference in means between group 0 and group 1 is not equal to 0
95 percent confidence interval:
-0.12927029 -0.02021856
sample estimates:
mean in group 0 mean in group 1
-0.03707873 0.03766570 t.test(fund.Bilat_NAcc_mask~Rel_p1Y_pos, data=reg_df,var.equal=TRUE)
Two Sample t-test
data: fund.Bilat_NAcc_mask by Rel_p1Y_pos
t = -1.6219, df = 42, p-value = 0.1123
alternative hypothesis: true difference in means between group 0 and group 1 is not equal to 0
95 percent confidence interval:
-0.086015229 0.009360935
sample estimates:
mean in group 0 mean in group 1
-0.01902793 0.01929922 t.test(rela.Bilat_NAcc_mask~Rel_p1Y_pos, data=reg_df,var.equal=TRUE)
Two Sample t-test
data: rela.Bilat_NAcc_mask by Rel_p1Y_pos
t = -1.1214, df = 42, p-value = 0.2685
alternative hypothesis: true difference in means between group 0 and group 1 is not equal to 0
95 percent confidence interval:
-0.06665278 0.01903601
sample estimates:
mean in group 0 mean in group 1
-0.01178564 0.01202275 t.test(news.Bilat_NAcc_mask~Rel_p1Y_pos, data=reg_df,var.equal=TRUE)
Two Sample t-test
data: news.Bilat_NAcc_mask by Rel_p1Y_pos
t = -0.40911, df = 42, p-value = 0.6845
alternative hypothesis: true difference in means between group 0 and group 1 is not equal to 0
95 percent confidence interval:
-0.05585200 0.03702392
sample estimates:
mean in group 0 mean in group 1
-0.004480793 0.004933247 Create data frame for NAcc plot
std <- function(x) sd(x)/sqrt(length(x))
x=Case_mean[c(19,9,4,24,14,35)]
AvgBeta_group = x %>% group_by(Performance = Rel_p1Y_pos) %>%
summarise("Company Profile" = mean(prof.Bilat_NAcc_mask),
"Price Graph" = mean(graph.Bilat_NAcc_mask),
"Fundamentals" = mean(fund.Bilat_NAcc_mask),
"Relative Valuation" = mean(rela.Bilat_NAcc_mask),
"News Item" = mean(news.Bilat_NAcc_mask))
SE_group = x %>% group_by(Performance = Rel_p1Y_pos) %>%
summarise("Company Profile" = std(prof.Bilat_NAcc_mask),
"Price Graph" = std(graph.Bilat_NAcc_mask),
"Fundamentals" = std(fund.Bilat_NAcc_mask),
"Relative Valuation" = std(rela.Bilat_NAcc_mask),
"News Item" = std(news.Bilat_NAcc_mask))
# Make Average dataframe
AvgBeta=AvgBeta_group
AvgBeta <- pivot_longer(AvgBeta, cols=-Performance,names_to='Screen',values_to='AvgBeta')
# Make SE dataframe
SE=SE_group
SE <- pivot_longer(SE, cols=-Performance,names_to='Screen',values_to='SE')
# Combine Average and SE
NAccplot_df <- AvgBeta %>% right_join(SE, by=c("Performance","Screen"))
#NAccplot_df
Create NAccplot
NAccplot = NAccplot_df
NAccplot$Performance <- factor(NAccplot$Performance)
NAccplot$Screen <- factor(NAccplot$Screen, levels=unique(NAccplot$Screen), ordered=TRUE)
plot <- ggplot(NAccplot, aes(x=Screen, y=AvgBeta, fill=Performance)) +
geom_bar(stat='identity', position = position_dodge2(padding = 0)) +
geom_errorbar(aes(ymin=AvgBeta-SE, ymax=AvgBeta+SE), width=.2, position=position_dodge(0.9))+
geom_hline(yintercept=0)+
labs(x = "Information screen", fill="Future stock performance",y=expression(paste("Average ", beta, "-estimate (centered)")))+
scale_fill_manual(labels = c("Underperforming", "Overperforming"), values = c("#8D8D8D", "#2070B4"))+
annotate("text", x = 1, y = 0.06, label ="paste(italic(p), \" = 0.005\")",parse=TRUE,size=6)+ # p-value from t-test above
annotate("text", x = 2, y = 0.06, label ="paste(italic(p), \" = 0.008\")",parse=TRUE,size=6)+ # p-value from t-test above
theme_classic()+
scale_y_continuous(expand = expansion(mult = c(0, 0)),limits = c(-0.07, 0.07),breaks=c(-0.06,-0.04,-0.02,0,0.02,0.04,0.06)) +
theme(legend.position="bottom", legend.box = "horizontal",text = element_text(size = 28,family="Arial"),
plot.title = element_text(hjust = 0.5))
#1. Open jpeg file
jpeg("NAcc_plot.jpg", width = 1200, height = 600)
#2. Create the plot
plot
#3. Close the file
dev.off()null device
1 #Create stock performance inflection variable
reg_df$Inflection <- ifelse(reg_df$Rel_m3Y_pos==reg_df$Rel_p1Y_pos, 0, 1)
formula=paste("Inflection~log_CMC+prof.Bilat_NAcc_mask+prof.Bilat_MPFC_mask+prof.Bilat_AIns_mask",sep="")
print(formula)[1] "Inflection~log_CMC+prof.Bilat_NAcc_mask+prof.Bilat_MPFC_mask+prof.Bilat_AIns_mask"out_prof_inflection <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_prof_inflection,digits = getOption("jtools-digits", default = 3)))MODEL INFO:
Observations: 44
Dependent Variable: Inflection
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(4) = 6.035, p = 0.197
Pseudo-R² (Cragg-Uhler) = 0.171
Pseudo-R² (McFadden) = 0.099
AIC = 64.962, BIC = 73.883
Standard errors: MLE
------------------------------------------------------------
Est. S.E. z val. p
-------------------------- -------- ------- -------- -------
(Intercept) 1.352 2.142 0.631 0.528
log_CMC -0.159 0.248 -0.639 0.523
prof.Bilat_NAcc_mask 6.480 5.684 1.140 0.254
prof.Bilat_MPFC_mask -9.869 4.736 -2.084 0.037
prof.Bilat_AIns_mask 11.286 7.334 1.539 0.124
------------------------------------------------------------print("No relationship between MPFC activity at profile screen and inflection:")[1] "No relationship between MPFC activity at profile screen and inflection:"formula=paste("Inflection~prof.Bilat_MPFC_mask",sep="")
print(formula)[1] "Inflection~prof.Bilat_MPFC_mask"out_prof_mpfc_inflection <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_prof_mpfc_inflection,digits = getOption("jtools-digits", default = 3)))MODEL INFO:
Observations: 44
Dependent Variable: Inflection
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(1) = 0.178, p = 0.673
Pseudo-R² (Cragg-Uhler) = 0.005
Pseudo-R² (McFadden) = 0.003
AIC = 64.819, BIC = 68.387
Standard errors: MLE
------------------------------------------------------------
Est. S.E. z val. p
-------------------------- -------- ------- -------- -------
(Intercept) 0.000 0.302 0.000 1.000
prof.Bilat_MPFC_mask -1.002 2.377 -0.421 0.673
------------------------------------------------------------formula=paste("Inflection~graph.Bilat_NAcc_mask+graph.Bilat_MPFC_mask+graph.Bilat_AIns_mask",sep="")
print(formula)[1] "Inflection~graph.Bilat_NAcc_mask+graph.Bilat_MPFC_mask+graph.Bilat_AIns_mask"out_graph_inflection <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_graph_inflection,digits = getOption("jtools-digits", default = 3)))MODEL INFO:
Observations: 44
Dependent Variable: Inflection
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(3) = 1.274, p = 0.735
Pseudo-R² (Cragg-Uhler) = 0.038
Pseudo-R² (McFadden) = 0.021
AIC = 67.723, BIC = 74.860
Standard errors: MLE
-------------------------------------------------------------
Est. S.E. z val. p
--------------------------- -------- ------- -------- -------
(Intercept) 0.001 0.306 0.002 0.999
graph.Bilat_NAcc_mask 2.497 5.860 0.426 0.670
graph.Bilat_MPFC_mask 2.701 3.508 0.770 0.441
graph.Bilat_AIns_mask -3.421 7.584 -0.451 0.652
-------------------------------------------------------------formula=paste("Inflection~fund.Bilat_NAcc_mask+fund.Bilat_MPFC_mask+fund.Bilat_AIns_mask",sep="")
print(formula)[1] "Inflection~fund.Bilat_NAcc_mask+fund.Bilat_MPFC_mask+fund.Bilat_AIns_mask"out_fund_inflection <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_fund_inflection,digits = getOption("jtools-digits", default = 3)))MODEL INFO:
Observations: 44
Dependent Variable: Inflection
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(3) = 2.863, p = 0.413
Pseudo-R² (Cragg-Uhler) = 0.084
Pseudo-R² (McFadden) = 0.047
AIC = 66.134, BIC = 73.270
Standard errors: MLE
-------------------------------------------------------------
Est. S.E. z val. p
-------------------------- -------- -------- -------- -------
(Intercept) 0.001 0.312 0.003 0.998
fund.Bilat_NAcc_mask -5.686 7.394 -0.769 0.442
fund.Bilat_MPFC_mask -3.191 3.844 -0.830 0.407
fund.Bilat_AIns_mask 16.939 11.142 1.520 0.128
-------------------------------------------------------------formula=paste("Inflection~rela.Bilat_NAcc_mask+rela.Bilat_MPFC_mask+rela.Bilat_AIns_mask",sep="")
print(formula)[1] "Inflection~rela.Bilat_NAcc_mask+rela.Bilat_MPFC_mask+rela.Bilat_AIns_mask"out_rela_inflection <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_rela_inflection,digits = getOption("jtools-digits", default = 3)))MODEL INFO:
Observations: 44
Dependent Variable: Inflection
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(3) = 1.350, p = 0.717
Pseudo-R² (Cragg-Uhler) = 0.040
Pseudo-R² (McFadden) = 0.022
AIC = 67.647, BIC = 74.783
Standard errors: MLE
------------------------------------------------------------
Est. S.E. z val. p
-------------------------- -------- ------- -------- -------
(Intercept) 0.002 0.306 0.006 0.995
rela.Bilat_NAcc_mask 5.972 7.528 0.793 0.428
rela.Bilat_MPFC_mask -3.913 4.005 -0.977 0.329
rela.Bilat_AIns_mask 1.866 8.161 0.229 0.819
------------------------------------------------------------formula=paste("Inflection~news.Bilat_NAcc_mask+news.Bilat_MPFC_mask+news.Bilat_AIns_mask",sep="")
print(formula)[1] "Inflection~news.Bilat_NAcc_mask+news.Bilat_MPFC_mask+news.Bilat_AIns_mask"out_news_inflection <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_news_inflection,digits = getOption("jtools-digits", default = 3)))MODEL INFO:
Observations: 44
Dependent Variable: Inflection
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(3) = 4.423, p = 0.219
Pseudo-R² (Cragg-Uhler) = 0.128
Pseudo-R² (McFadden) = 0.073
AIC = 64.574, BIC = 71.711
Standard errors: MLE
------------------------------------------------------------
Est. S.E. z val. p
-------------------------- -------- ------- -------- -------
(Intercept) -0.003 0.317 -0.008 0.994
news.Bilat_NAcc_mask 10.310 6.769 1.523 0.128
news.Bilat_MPFC_mask -8.348 4.549 -1.835 0.066
news.Bilat_AIns_mask 2.711 6.771 0.400 0.689
------------------------------------------------------------print("No relationship between MPFC activity at news screen and inflection:")[1] "No relationship between MPFC activity at news screen and inflection:"formula=paste("Inflection~news.Bilat_MPFC_mask",sep="")
print(formula)[1] "Inflection~news.Bilat_MPFC_mask"out_news_mpfc_inflection <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_news_mpfc_inflection,digits = getOption("jtools-digits", default = 3)))MODEL INFO:
Observations: 44
Dependent Variable: Inflection
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(1) = 0.615, p = 0.433
Pseudo-R² (Cragg-Uhler) = 0.019
Pseudo-R² (McFadden) = 0.010
AIC = 64.382, BIC = 67.950
Standard errors: MLE
------------------------------------------------------------
Est. S.E. z val. p
-------------------------- -------- ------- -------- -------
(Intercept) 0.001 0.304 0.004 0.997
news.Bilat_MPFC_mask -2.221 2.861 -0.776 0.438
------------------------------------------------------------formula=paste("Rel_p1Y_pos~Factor1",sep="")
print(formula)[1] "Rel_p1Y_pos~Factor1"out_factor1 <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_factor1,digits = getOption("jtools-digits", default = 3),scale=TRUE))MODEL INFO:
Observations: 44
Dependent Variable: Rel_p1Y_pos
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(1) = 7.100, p = 0.008
Pseudo-R² (Cragg-Uhler) = 0.199
Pseudo-R² (McFadden) = 0.116
AIC = 57.897, BIC = 61.465
Standard errors: MLE
--------------------------------------------------
Est. S.E. z val. p
----------------- ------- ------- -------- -------
(Intercept) 0.010 0.327 0.030 0.976
Factor1 0.913 0.382 2.388 0.017
--------------------------------------------------
Continuous predictors are mean-centered and scaled by 1 s.d. The outcome variable remains in its original units.formula=paste("Rel_p1Y_pos~Factor2",sep="")
print(formula)[1] "Rel_p1Y_pos~Factor2"out_factor2 <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_factor2,digits = getOption("jtools-digits", default = 3),scale=TRUE))MODEL INFO:
Observations: 44
Dependent Variable: Rel_p1Y_pos
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(1) = 0.113, p = 0.737
Pseudo-R² (Cragg-Uhler) = 0.003
Pseudo-R² (McFadden) = 0.002
AIC = 64.884, BIC = 68.452
Standard errors: MLE
---------------------------------------------------
Est. S.E. z val. p
----------------- -------- ------- -------- -------
(Intercept) -0.000 0.302 -0.000 1.000
Factor2 -0.103 0.307 -0.335 0.737
---------------------------------------------------
Continuous predictors are mean-centered and scaled by 1 s.d. The outcome variable remains in its original units.formula=paste("Rel_p1Y_pos~log_CMC+Rel_m3Y_pos+f_s+f_ebit+f_ebitmargin+f_roe+f_roic+rv_evs+rv_ebitda+rv_pe+rv_pb+news_incgr",sep="")
print(formula)[1] "Rel_p1Y_pos~log_CMC+Rel_m3Y_pos+f_s+f_ebit+f_ebitmargin+f_roe+f_roic+rv_evs+rv_ebitda+rv_pe+rv_pb+news_incgr"out_market <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_market,digits = getOption("jtools-digits", default = 3),scale=TRUE))MODEL INFO:
Observations: 44
Dependent Variable: Rel_p1Y_pos
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(12) = 13.099, p = 0.362
Pseudo-R² (Cragg-Uhler) = 0.343
Pseudo-R² (McFadden) = 0.215
AIC = 73.898, BIC = 97.093
Standard errors: MLE
----------------------------------------------------
Est. S.E. z val. p
------------------ -------- ------- -------- -------
(Intercept) 0.309 0.813 0.380 0.704
log_CMC 0.751 0.766 0.981 0.327
Rel_m3Y_pos 0.194 0.847 0.229 0.819
f_s -0.360 0.485 -0.743 0.457
f_ebit -0.294 0.577 -0.510 0.610
f_ebitmargin -0.147 0.731 -0.201 0.841
f_roe -0.408 0.940 -0.434 0.664
f_roic -1.005 0.833 -1.207 0.228
rv_evs 0.004 0.887 0.004 0.997
rv_ebitda 1.989 1.210 1.644 0.100
rv_pe -1.225 0.775 -1.581 0.114
rv_pb 1.252 3.684 0.340 0.734
news_incgr 1.267 1.520 0.834 0.404
----------------------------------------------------
Continuous predictors are mean-centered and scaled by 1 s.d. The outcome variable remains in its original units.formula=paste("Rel_p1Y_pos~log_CMC+Rel_m3Y_pos+f_s+f_ebit+f_ebitmargin+f_roe+f_roic+rv_evs+rv_ebitda+rv_pe+rv_pb+news_incgr+choice",sep="")
print(formula)[1] "Rel_p1Y_pos~log_CMC+Rel_m3Y_pos+f_s+f_ebit+f_ebitmargin+f_roe+f_roic+rv_evs+rv_ebitda+rv_pe+rv_pb+news_incgr+choice"out_market_choice <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_market_choice,digits = getOption("jtools-digits", default = 3),scale=TRUE))MODEL INFO:
Observations: 44
Dependent Variable: Rel_p1Y_pos
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(13) = 14.440, p = 0.344
Pseudo-R² (Cragg-Uhler) = 0.373
Pseudo-R² (McFadden) = 0.237
AIC = 74.557, BIC = 99.535
Standard errors: MLE
----------------------------------------------------
Est. S.E. z val. p
------------------ -------- ------- -------- -------
(Intercept) 0.285 0.789 0.362 0.718
log_CMC 0.864 0.763 1.132 0.258
Rel_m3Y_pos -0.068 0.896 -0.076 0.939
f_s -0.204 0.498 -0.410 0.682
f_ebit -0.577 0.616 -0.938 0.348
f_ebitmargin -0.035 0.796 -0.044 0.965
f_roe -0.370 0.961 -0.386 0.700
f_roic -1.298 0.971 -1.336 0.182
rv_evs 0.047 0.893 0.053 0.958
rv_ebitda 1.980 1.243 1.593 0.111
rv_pe -1.256 0.724 -1.735 0.083
rv_pb 1.032 3.297 0.313 0.754
news_incgr 0.547 1.568 0.349 0.727
choice 0.565 0.500 1.131 0.258
----------------------------------------------------
Continuous predictors are mean-centered and scaled by 1 s.d. The outcome variable remains in its original units.formula=paste("Rel_p1Y_pos~log_CMC+Rel_m3Y_pos+f_s+f_ebit+f_ebitmargin+f_roe+f_roic+rv_evs+rv_ebitda+rv_pe+rv_pb+news_incgr+choice+Factor1",sep="")
print(formula)[1] "Rel_p1Y_pos~log_CMC+Rel_m3Y_pos+f_s+f_ebit+f_ebitmargin+f_roe+f_roic+rv_evs+rv_ebitda+rv_pe+rv_pb+news_incgr+choice+Factor1"out_all <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_all,digits = getOption("jtools-digits", default = 3),scale=TRUE))MODEL INFO:
Observations: 44
Dependent Variable: Rel_p1Y_pos
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(14) = 25.334, p = 0.031
Pseudo-R² (Cragg-Uhler) = 0.584
Pseudo-R² (McFadden) = 0.415
AIC = 65.663, BIC = 92.426
Standard errors: MLE
----------------------------------------------------
Est. S.E. z val. p
------------------ -------- ------- -------- -------
(Intercept) 0.165 0.811 0.204 0.839
log_CMC 1.195 1.082 1.105 0.269
Rel_m3Y_pos 0.357 1.083 0.330 0.742
f_s 0.934 0.741 1.261 0.207
f_ebit -1.374 0.920 -1.494 0.135
f_ebitmargin 0.388 1.107 0.350 0.726
f_roe -1.149 1.577 -0.729 0.466
f_roic -1.037 1.122 -0.924 0.356
rv_evs 0.986 1.010 0.976 0.329
rv_ebitda 2.539 1.495 1.699 0.089
rv_pe -1.844 0.911 -2.024 0.043
rv_pb 0.199 1.176 0.169 0.866
news_incgr 2.326 1.866 1.247 0.212
choice 1.299 0.764 1.699 0.089
Factor1 1.879 0.741 2.534 0.011
----------------------------------------------------
Continuous predictors are mean-centered and scaled by 1 s.d. The outcome variable remains in its original units.Create Table 3:
coef_names <- c("Current Market Capitalization" = "log_CMC",
"36M Performance"='Rel_m3Y_pos',
"Sales" ="f_s", "EBIT" = "f_ebit", "EBIT Margin (%)" = "f_ebitmargin","ROE (%)" = "f_roe","ROIC (%)" = "f_roic",
"EV/Sales" = "rv_evs","EV/EBITDA" = "rv_ebitda","P/E" = "rv_pe","P/B" = "rv_pb",
"Income growth" = "news_incgr","Prediction" = "choice","NAcc - Profile, Graph" = "Factor1")
# Uncomment to create table
#quick_docx(export_summs(out_market,out_market_choice,out_all, scale = TRUE,coefs = coef_names,error_pos = c("same"),
# model.names=c("Market","Market + Behavior","Market + Behavior + Brain"),
# statistics= c(N = "nobs", AIC = "AIC", "Pseudo R2 (McFadden)" = "pseudo.r.squared.mcfadden")),file='Table-3.docx')Control stock metric: rv_pe
formula=paste("Rel_p1Y_pos~rv_pe",sep="")
print(formula)[1] "Rel_p1Y_pos~rv_pe"out_rvpe <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_rvpe,digits = getOption("jtools-digits", default = 3),scale=TRUE))MODEL INFO:
Observations: 44
Dependent Variable: Rel_p1Y_pos
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(1) = 0.029, p = 0.866
Pseudo-R² (Cragg-Uhler) = 0.001
Pseudo-R² (McFadden) = 0.000
AIC = 64.968, BIC = 68.537
Standard errors: MLE
---------------------------------------------------
Est. S.E. z val. p
----------------- -------- ------- -------- -------
(Intercept) -0.000 0.302 -0.000 1.000
rv_pe -0.052 0.306 -0.169 0.866
---------------------------------------------------
Continuous predictors are mean-centered and scaled by 1 s.d. The outcome variable remains in its original units.Model comparison:
anova(out_market, out_all, test = "LRT")Analysis of Deviance Table
Model 1: Rel_p1Y_pos ~ log_CMC + Rel_m3Y_pos + f_s + f_ebit + f_ebitmargin +
f_roe + f_roic + rv_evs + rv_ebitda + rv_pe + rv_pb + news_incgr
Model 2: Rel_p1Y_pos ~ log_CMC + Rel_m3Y_pos + f_s + f_ebit + f_ebitmargin +
f_roe + f_roic + rv_evs + rv_ebitda + rv_pe + rv_pb + news_incgr +
choice + Factor1
Resid. Df Resid. Dev Df Deviance Pr(>Chi)
1 31 47.898
2 29 35.663 2 12.236 0.002203 **
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1anova(out_market_choice, out_all, test = "LRT")Analysis of Deviance Table
Model 1: Rel_p1Y_pos ~ log_CMC + Rel_m3Y_pos + f_s + f_ebit + f_ebitmargin +
f_roe + f_roic + rv_evs + rv_ebitda + rv_pe + rv_pb + news_incgr +
choice
Model 2: Rel_p1Y_pos ~ log_CMC + Rel_m3Y_pos + f_s + f_ebit + f_ebitmargin +
f_roe + f_roic + rv_evs + rv_ebitda + rv_pe + rv_pb + news_incgr +
choice + Factor1
Resid. Df Resid. Dev Df Deviance Pr(>Chi)
1 30 46.557
2 29 35.663 1 10.894 0.0009647 ***
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1Set input for prediction analysis:
crossval_df = reg_df[, c("Rel_p1Y_pos",
"choice",
"log_CMC","Rel_m3Y_pos","f_s","f_ebit","f_ebitmargin","f_roe","f_roic","rv_evs","rv_ebitda","rv_pe","rv_pb","news_incgr",
"Factor1")]
crossval_df$Rel_m3Y_pos <- as.factor(crossval_df$Rel_m3Y_pos)
crossval_df$Rel_p1Y_pos <- as.factor(crossval_df$Rel_p1Y_pos)
crossval_dfK-fold cross-validation with the caret package:
K-fold: Brain (NAcc)
braindat = crossval_df[,c("Rel_p1Y_pos","Factor1")]
choicedat = crossval_df[,c("Rel_p1Y_pos","choice")]
marketdat = crossval_df[,-c(2,15)]
set.seed(0)
kfold = function(data, nfolds){
folds <- createFolds(factor(data$Rel_p1Y_pos), nfolds, returnTrain = T)
train_model<- trainControl(method="cv", number=nfolds, index = folds, savePredictions = TRUE)
# Model
model <- train(
Rel_p1Y_pos~ .,
data = data,
method = "glm", # logistic regression
family = "binomial",
trControl = train_model)
return(confusionMatrix(model$pred$pred,model$pred$obs,positive="1"))
}
print(kfold(braindat,5))Confusion Matrix and Statistics
Reference
Prediction 0 1
0 16 8
1 6 14
Accuracy : 0.6818
95% CI : (0.5242, 0.8139)
No Information Rate : 0.5
P-Value [Acc > NIR] : 0.01131
Kappa : 0.3636
Mcnemar's Test P-Value : 0.78927
Sensitivity : 0.6364
Specificity : 0.7273
Pos Pred Value : 0.7000
Neg Pred Value : 0.6667
Prevalence : 0.5000
Detection Rate : 0.3182
Detection Prevalence : 0.4545
Balanced Accuracy : 0.6818
'Positive' Class : 1
K-fold: Behavior (choice)
print(kfold(choicedat,5))Confusion Matrix and Statistics
Reference
Prediction 0 1
0 10 13
1 12 9
Accuracy : 0.4318
95% CI : (0.2835, 0.5897)
No Information Rate : 0.5
P-Value [Acc > NIR] : 0.8544
Kappa : -0.1364
Mcnemar's Test P-Value : 1.0000
Sensitivity : 0.4091
Specificity : 0.4545
Pos Pred Value : 0.4286
Neg Pred Value : 0.4348
Prevalence : 0.5000
Detection Rate : 0.2045
Detection Prevalence : 0.4773
Balanced Accuracy : 0.4318
'Positive' Class : 1
K-fold: Market (Stock metrics)
print(kfold(marketdat,5))Confusion Matrix and Statistics
Reference
Prediction 0 1
0 8 11
1 14 11
Accuracy : 0.4318
95% CI : (0.2835, 0.5897)
No Information Rate : 0.5
P-Value [Acc > NIR] : 0.8544
Kappa : -0.1364
Mcnemar's Test P-Value : 0.6892
Sensitivity : 0.5000
Specificity : 0.3636
Pos Pred Value : 0.4400
Neg Pred Value : 0.4211
Prevalence : 0.5000
Detection Rate : 0.2500
Detection Prevalence : 0.5682
Balanced Accuracy : 0.4318
'Positive' Class : 1
Leave-one-out cross-validation analyses:
Leave-one-out: Brain
set.seed(0)
# LOOCV
train_model<- trainControl(method="loocv", savePredictions = TRUE)
# Model
model <- train(
Rel_p1Y_pos ~ .,
data = braindat,
method = "glm", # logistic regression
family = "binomial",
trControl = train_model)
model #to print model resultGeneralized Linear Model
44 samples
1 predictor
2 classes: '0', '1'
No pre-processing
Resampling: Leave-One-Out Cross-Validation
Summary of sample sizes: 43, 43, 43, 43, 43, 43, ...
Resampling results:
Accuracy Kappa
0.6363636 0 confusionMatrix(model$pred$pred,model$pred$obs,positive="1")Confusion Matrix and Statistics
Reference
Prediction 0 1
0 15 9
1 7 13
Accuracy : 0.6364
95% CI : (0.4777, 0.7759)
No Information Rate : 0.5
P-Value [Acc > NIR] : 0.04807
Kappa : 0.2727
Mcnemar's Test P-Value : 0.80259
Sensitivity : 0.5909
Specificity : 0.6818
Pos Pred Value : 0.6500
Neg Pred Value : 0.6250
Prevalence : 0.5000
Detection Rate : 0.2955
Detection Prevalence : 0.4545
Balanced Accuracy : 0.6364
'Positive' Class : 1
Leave-one-out: Behavior (choice)
set.seed(0)
# LOOCV
train_model<- trainControl(method="loocv", savePredictions = TRUE)
# Model
model <- train(
Rel_p1Y_pos ~ .,
data = choicedat,
method = "glm", # logistic regression
family = "binomial",
trControl = train_model)
model #to print model resultGeneralized Linear Model
44 samples
1 predictor
2 classes: '0', '1'
No pre-processing
Resampling: Leave-One-Out Cross-Validation
Summary of sample sizes: 43, 43, 43, 43, 43, 43, ...
Resampling results:
Accuracy Kappa
0.5 0 confusionMatrix(model$pred$pred,model$pred$obs,positive="1")Confusion Matrix and Statistics
Reference
Prediction 0 1
0 10 10
1 12 12
Accuracy : 0.5
95% CI : (0.3456, 0.6544)
No Information Rate : 0.5
P-Value [Acc > NIR] : 0.5598
Kappa : 0
Mcnemar's Test P-Value : 0.8312
Sensitivity : 0.5455
Specificity : 0.4545
Pos Pred Value : 0.5000
Neg Pred Value : 0.5000
Prevalence : 0.5000
Detection Rate : 0.2727
Detection Prevalence : 0.5455
Balanced Accuracy : 0.5000
'Positive' Class : 1
Leave-one-out: Market (Stock metrics)
set.seed(0)
# LOOCV
train_model<- trainControl(method="loocv", savePredictions = TRUE)
# Model
model <- train(
Rel_p1Y_pos ~ .,
data = marketdat,
method = "glm", # logistic regression
family = "binomial",
trControl = train_model)
model #to print model resultGeneralized Linear Model
44 samples
12 predictors
2 classes: '0', '1'
No pre-processing
Resampling: Leave-One-Out Cross-Validation
Summary of sample sizes: 43, 43, 43, 43, 43, 43, ...
Resampling results:
Accuracy Kappa
0.4318182 0 confusionMatrix(model$pred$pred,model$pred$obs,positive="1")Confusion Matrix and Statistics
Reference
Prediction 0 1
0 10 13
1 12 9
Accuracy : 0.4318
95% CI : (0.2835, 0.5897)
No Information Rate : 0.5
P-Value [Acc > NIR] : 0.8544
Kappa : -0.1364
Mcnemar's Test P-Value : 1.0000
Sensitivity : 0.4091
Specificity : 0.4545
Pos Pred Value : 0.4286
Neg Pred Value : 0.4348
Prevalence : 0.5000
Detection Rate : 0.2045
Detection Prevalence : 0.4773
Balanced Accuracy : 0.4318
'Positive' Class : 1
Table S1: Logistic regression model for control regions
formula=paste("Rel_p1Y_pos~prof.Left_PIns_mask+prof.Left_PVis_mask",sep="")
print(formula)[1] "Rel_p1Y_pos~prof.Left_PIns_mask+prof.Left_PVis_mask"out_prof_control <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_prof_control,digits = getOption("jtools-digits", default = 3),scale=TRUE))MODEL INFO:
Observations: 44
Dependent Variable: Rel_p1Y_pos
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(2) = 0.037, p = 0.982
Pseudo-R² (Cragg-Uhler) = 0.001
Pseudo-R² (McFadden) = 0.001
AIC = 66.960, BIC = 72.313
Standard errors: MLE
-----------------------------------------------------------
Est. S.E. z val. p
------------------------- -------- ------- -------- -------
(Intercept) -0.000 0.302 -0.000 1.000
prof.Left_PIns_mask -0.071 0.375 -0.191 0.849
prof.Left_PVis_mask 0.036 0.375 0.097 0.923
-----------------------------------------------------------
Continuous predictors are mean-centered and scaled by 1 s.d. The outcome variable remains in its original units.formula=paste("Rel_p1Y_pos~graph.Left_PIns_mask+graph.Left_PVis_mask",sep="")
print(formula)[1] "Rel_p1Y_pos~graph.Left_PIns_mask+graph.Left_PVis_mask"out_graph_control <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_graph_control,digits = getOption("jtools-digits", default = 3),scale=TRUE))MODEL INFO:
Observations: 44
Dependent Variable: Rel_p1Y_pos
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(2) = 2.846, p = 0.241
Pseudo-R² (Cragg-Uhler) = 0.084
Pseudo-R² (McFadden) = 0.047
AIC = 64.151, BIC = 69.504
Standard errors: MLE
------------------------------------------------------------
Est. S.E. z val. p
-------------------------- -------- ------- -------- -------
(Intercept) 0.001 0.311 0.002 0.999
graph.Left_PIns_mask -0.159 0.412 -0.387 0.699
graph.Left_PVis_mask 0.628 0.429 1.461 0.144
------------------------------------------------------------
Continuous predictors are mean-centered and scaled by 1 s.d. The outcome variable remains in its original units.formula=paste("Rel_p1Y_pos~fund.Left_PIns_mask+fund.Left_PVis_mask",sep="")
print(formula)[1] "Rel_p1Y_pos~fund.Left_PIns_mask+fund.Left_PVis_mask"out_fund_control <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_fund_control,digits = getOption("jtools-digits", default = 3),scale=TRUE))MODEL INFO:
Observations: 44
Dependent Variable: Rel_p1Y_pos
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(2) = 4.368, p = 0.113
Pseudo-R² (Cragg-Uhler) = 0.126
Pseudo-R² (McFadden) = 0.072
AIC = 62.629, BIC = 67.982
Standard errors: MLE
-----------------------------------------------------------
Est. S.E. z val. p
------------------------- -------- ------- -------- -------
(Intercept) 0.003 0.317 0.011 0.991
fund.Left_PIns_mask -0.728 0.423 -1.722 0.085
fund.Left_PVis_mask 0.816 0.450 1.813 0.070
-----------------------------------------------------------
Continuous predictors are mean-centered and scaled by 1 s.d. The outcome variable remains in its original units.formula=paste("Rel_p1Y_pos~rela.Left_PIns_mask+rela.Left_PVis_mask",sep="")
print(formula)[1] "Rel_p1Y_pos~rela.Left_PIns_mask+rela.Left_PVis_mask"out_rela_control <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_rela_control,digits = getOption("jtools-digits", default = 3),scale=TRUE))MODEL INFO:
Observations: 44
Dependent Variable: Rel_p1Y_pos
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(2) = 0.533, p = 0.766
Pseudo-R² (Cragg-Uhler) = 0.016
Pseudo-R² (McFadden) = 0.009
AIC = 66.464, BIC = 71.817
Standard errors: MLE
-----------------------------------------------------------
Est. S.E. z val. p
------------------------- -------- ------- -------- -------
(Intercept) -0.000 0.303 -0.001 0.999
rela.Left_PIns_mask -0.104 0.402 -0.259 0.795
rela.Left_PVis_mask 0.278 0.409 0.680 0.497
-----------------------------------------------------------
Continuous predictors are mean-centered and scaled by 1 s.d. The outcome variable remains in its original units.formula=paste("Rel_p1Y_pos~news.Left_PIns_mask+news.Left_PVis_mask",sep="")
print(formula)[1] "Rel_p1Y_pos~news.Left_PIns_mask+news.Left_PVis_mask"out_news_control <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_news_control,digits = getOption("jtools-digits", default = 3),scale=TRUE))MODEL INFO:
Observations: 44
Dependent Variable: Rel_p1Y_pos
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(2) = 0.085, p = 0.958
Pseudo-R² (Cragg-Uhler) = 0.003
Pseudo-R² (McFadden) = 0.001
AIC = 66.912, BIC = 72.264
Standard errors: MLE
----------------------------------------------------------
Est. S.E. z val. p
------------------------- ------- ------- -------- -------
(Intercept) 0.000 0.302 0.000 1.000
news.Left_PIns_mask 0.019 0.339 0.055 0.956
news.Left_PVis_mask 0.079 0.340 0.234 0.815
----------------------------------------------------------
Continuous predictors are mean-centered and scaled by 1 s.d. The outcome variable remains in its original units.coef_names <- c("Profile - PVis" = "prof.Left_PVis_mask","Profile - PIns" = "prof.Left_PIns_mask",
"Graph - PVis" = "graph.Left_PVis_mask","Graph - PIns" = "graph.Left_PIns_mask",
"Fundamentals - PVis" = "fund.Left_PVis_mask","Fundamentals - PIns" = "fund.Left_PIns_mask",
"Relative Valuation - PVis" = "rela.Left_PVis_mask","Relative Valuation - PIns" = "rela.Left_PIns_mask",
"News Item - PVis" = "news.Left_PVis_mask","News Item - PIns" = "news.Left_PIns_mask")
# Uncomment to create table
#quick_docx(export_summs(out_prof_control,out_graph_control,out_fund_control,out_rela_control,out_news_control, scale = TRUE,coefs = coef_names,error_pos = c("same"),
# model.names=c("Company Profile","Price Graph","Fundamentals","Relative Valuation","News Item"),
# statistics= c(N = "nobs", AIC = "AIC", "Pseudo R2 (McFadden)" = "pseudo.r.squared.mcfadden")), file='Table-S1.docx')Table S2: Correlation matrix:
prof_data <- reg_df[, c(18,19,20,21,22)]
names(prof_data) <- substring(names(prof_data), 6)
graph_data <- reg_df[, c(8,9,10,11,12)]
names(graph_data) <- substring(names(graph_data), 7)
fund_data <- reg_df[, c(3,4,5,6,7)]
names(fund_data) <- substring(names(fund_data), 6)
rela_data <- reg_df[, c(23,24,25,26,27)]
names(rela_data) <- substring(names(rela_data), 6)
news_data <- reg_df[, c(13,14,15,16,17)]
names(news_data) <- substring(names(news_data), 6)
res_prof <- cor(prof_data)
res_graph <-cor(graph_data)
res_fund <-cor(fund_data)
res_rela <-cor(rela_data)
res_news <-cor(news_data)
cor_matrices <-list(res_prof,res_graph,res_fund,res_rela,res_news)
Reduce("+", cor_matrices) / length(cor_matrices) Bilat_MPFC_mask Bilat_AIns_mask Bilat_NAcc_mask Left_PVis_mask Left_PIns_mask
Bilat_MPFC_mask 1.0000000 0.7104264 0.6892672 0.4125912 0.4929137
Bilat_AIns_mask 0.7104264 1.0000000 0.7615776 0.4339454 0.4942590
Bilat_NAcc_mask 0.6892672 0.7615776 1.0000000 0.5142567 0.5206732
Left_PVis_mask 0.4125912 0.4339454 0.5142567 1.0000000 0.5825221
Left_PIns_mask 0.4929137 0.4942590 0.5206732 0.5825221 1.0000000Table S3: VOIs and Control regions
formula=paste("Rel_p1Y_pos~prof.Bilat_NAcc_mask+prof.Bilat_MPFC_mask+prof.Bilat_AIns_mask+prof.Left_PIns_mask+prof.Left_PVis_mask",sep="")
print(formula)[1] "Rel_p1Y_pos~prof.Bilat_NAcc_mask+prof.Bilat_MPFC_mask+prof.Bilat_AIns_mask+prof.Left_PIns_mask+prof.Left_PVis_mask"out_prof_control2 <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_prof_control2,digits = getOption("jtools-digits", default = 3),scale=TRUE))MODEL INFO:
Observations: 44
Dependent Variable: Rel_p1Y_pos
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(5) = 12.224, p = 0.032
Pseudo-R² (Cragg-Uhler) = 0.323
Pseudo-R² (McFadden) = 0.200
AIC = 60.773, BIC = 71.478
Standard errors: MLE
------------------------------------------------------------
Est. S.E. z val. p
-------------------------- -------- ------- -------- -------
(Intercept) 0.028 0.352 0.079 0.937
prof.Bilat_NAcc_mask 1.634 0.683 2.393 0.017
prof.Bilat_MPFC_mask 0.186 0.587 0.317 0.751
prof.Bilat_AIns_mask -0.583 0.576 -1.013 0.311
prof.Left_PIns_mask 0.007 0.452 0.015 0.988
prof.Left_PVis_mask -0.634 0.478 -1.327 0.185
------------------------------------------------------------
Continuous predictors are mean-centered and scaled by 1 s.d. The outcome variable remains in its original units.formula=paste("Rel_p1Y_pos~graph.Bilat_NAcc_mask+graph.Bilat_MPFC_mask+graph.Bilat_AIns_mask+graph.Left_PIns_mask+graph.Left_PVis_mask",sep="")
print(formula)[1] "Rel_p1Y_pos~graph.Bilat_NAcc_mask+graph.Bilat_MPFC_mask+graph.Bilat_AIns_mask+graph.Left_PIns_mask+graph.Left_PVis_mask"out_graph_control2 <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_graph_control2,digits = getOption("jtools-digits", default = 3),scale=TRUE))MODEL INFO:
Observations: 44
Dependent Variable: Rel_p1Y_pos
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(5) = 11.423, p = 0.044
Pseudo-R² (Cragg-Uhler) = 0.305
Pseudo-R² (McFadden) = 0.187
AIC = 61.573, BIC = 72.279
Standard errors: MLE
-------------------------------------------------------------
Est. S.E. z val. p
--------------------------- -------- ------- -------- -------
(Intercept) -0.011 0.347 -0.033 0.974
graph.Bilat_NAcc_mask 1.784 0.808 2.208 0.027
graph.Bilat_MPFC_mask -0.835 0.589 -1.418 0.156
graph.Bilat_AIns_mask -0.222 0.730 -0.304 0.761
graph.Left_PIns_mask -0.249 0.467 -0.535 0.593
graph.Left_PVis_mask 0.258 0.522 0.494 0.621
-------------------------------------------------------------
Continuous predictors are mean-centered and scaled by 1 s.d. The outcome variable remains in its original units.formula=paste("Rel_p1Y_pos~fund.Bilat_NAcc_mask+fund.Bilat_MPFC_mask+fund.Bilat_AIns_mask+fund.Left_PIns_mask+fund.Left_PVis_mask",sep="")
print(formula)[1] "Rel_p1Y_pos~fund.Bilat_NAcc_mask+fund.Bilat_MPFC_mask+fund.Bilat_AIns_mask+fund.Left_PIns_mask+fund.Left_PVis_mask"out_fund_control2 <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_fund_control2,digits = getOption("jtools-digits", default = 3),scale=TRUE))MODEL INFO:
Observations: 44
Dependent Variable: Rel_p1Y_pos
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(5) = 12.780, p = 0.026
Pseudo-R² (Cragg-Uhler) = 0.336
Pseudo-R² (McFadden) = 0.210
AIC = 60.217, BIC = 70.922
Standard errors: MLE
------------------------------------------------------------
Est. S.E. z val. p
-------------------------- -------- ------- -------- -------
(Intercept) -0.043 0.353 -0.123 0.902
fund.Bilat_NAcc_mask 2.114 0.880 2.403 0.016
fund.Bilat_MPFC_mask 0.401 0.560 0.716 0.474
fund.Bilat_AIns_mask -1.521 0.849 -1.792 0.073
fund.Left_PIns_mask -1.333 0.554 -2.406 0.016
fund.Left_PVis_mask 0.689 0.492 1.399 0.162
------------------------------------------------------------
Continuous predictors are mean-centered and scaled by 1 s.d. The outcome variable remains in its original units.formula=paste("Rel_p1Y_pos~rela.Bilat_NAcc_mask+rela.Bilat_MPFC_mask+rela.Bilat_AIns_mask+rela.Left_PIns_mask+rela.Left_PVis_mask",sep="")
print(formula)[1] "Rel_p1Y_pos~rela.Bilat_NAcc_mask+rela.Bilat_MPFC_mask+rela.Bilat_AIns_mask+rela.Left_PIns_mask+rela.Left_PVis_mask"out_rela_control2 <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_rela_control2,digits = getOption("jtools-digits", default = 3),scale=TRUE))MODEL INFO:
Observations: 44
Dependent Variable: Rel_p1Y_pos
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(5) = 5.191, p = 0.393
Pseudo-R² (Cragg-Uhler) = 0.148
Pseudo-R² (McFadden) = 0.085
AIC = 67.806, BIC = 78.511
Standard errors: MLE
------------------------------------------------------------
Est. S.E. z val. p
-------------------------- -------- ------- -------- -------
(Intercept) -0.007 0.320 -0.023 0.982
rela.Bilat_NAcc_mask 0.949 0.603 1.575 0.115
rela.Bilat_MPFC_mask 0.648 0.566 1.144 0.253
rela.Bilat_AIns_mask -0.888 0.574 -1.545 0.122
rela.Left_PIns_mask -0.625 0.565 -1.106 0.269
rela.Left_PVis_mask 0.172 0.427 0.404 0.687
------------------------------------------------------------
Continuous predictors are mean-centered and scaled by 1 s.d. The outcome variable remains in its original units.formula=paste("Rel_p1Y_pos~news.Bilat_NAcc_mask+news.Bilat_MPFC_mask+news.Bilat_AIns_mask+news.Left_PIns_mask+news.Left_PVis_mask",sep="")
print(formula)[1] "Rel_p1Y_pos~news.Bilat_NAcc_mask+news.Bilat_MPFC_mask+news.Bilat_AIns_mask+news.Left_PIns_mask+news.Left_PVis_mask"out_news_control2 <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_news_control2,digits = getOption("jtools-digits", default = 3),scale=TRUE))MODEL INFO:
Observations: 44
Dependent Variable: Rel_p1Y_pos
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(5) = 3.080, p = 0.688
Pseudo-R² (Cragg-Uhler) = 0.090
Pseudo-R² (McFadden) = 0.050
AIC = 69.917, BIC = 80.622
Standard errors: MLE
------------------------------------------------------------
Est. S.E. z val. p
-------------------------- -------- ------- -------- -------
(Intercept) -0.003 0.312 -0.008 0.994
news.Bilat_NAcc_mask -0.345 0.505 -0.683 0.495
news.Bilat_MPFC_mask 0.415 0.498 0.835 0.404
news.Bilat_AIns_mask 0.573 0.472 1.214 0.225
news.Left_PIns_mask -0.310 0.437 -0.710 0.478
news.Left_PVis_mask 0.018 0.368 0.049 0.961
------------------------------------------------------------
Continuous predictors are mean-centered and scaled by 1 s.d. The outcome variable remains in its original units.coef_names <- c("Profile - PVis" = "prof.Left_PVis_mask","Profile - PIns" = "prof.Left_PIns_mask","Profile - NAcc" = "prof.Bilat_NAcc_mask", "Profile - MPFC" = "prof.Bilat_MPFC_mask", "Profile - AIns" = "prof.Bilat_AIns_mask",
"Graph - PVis" = "graph.Left_PVis_mask","Graph - PIns" = "graph.Left_PIns_mask","Graph - NAcc" = "graph.Bilat_NAcc_mask", "Graph - MPFC" = "graph.Bilat_MPFC_mask", "Graph - AIns" = "graph.Bilat_AIns_mask",
"Fundamentals - PVis" = "fund.Left_PVis_mask","Fundamentals - PIns" = "fund.Left_PIns_mask","Fundamentals - NAcc" = "fund.Bilat_NAcc_mask", "Fundamentals - MPFC" = "fund.Bilat_MPFC_mask", "Fundamentals - AIns" = "fund.Bilat_AIns_mask",
"Relative Valuation - PVis" = "rela.Left_PVis_mask","Relative Valuation - PIns" = "rela.Left_PIns_mask","Relative Valuation - NAcc" = "rela.Bilat_NAcc_mask", "Relative Valuation - MPFC" = "rela.Bilat_MPFC_mask", "Relative Valuation - AIns" = "rela.Bilat_AIns_mask",
"News item - PVis" = "news.Left_PVis_mask","News item - PIns" = "news.Left_PIns_mask","News item - NAcc" = "news.Bilat_NAcc_mask", "News item - MPFC" = "news.Bilat_MPFC_mask", "News item - AIns" = "news.Bilat_AIns_mask" )
# Uncomment to create table
#quick_docx(export_summs(out_prof_control2,out_graph_control2,out_fund_control2,out_rela_control2,out_news_control2, scale = TRUE,coefs = coef_names,error_pos = c("same"),
# model.names=c("Company Profile","Price Graph","Fundamentals","Relative Valuation","News Item"),
# statistics= c(N = "nobs", AIC = "AIC", "Pseudo R2 (McFadden)" = "pseudo.r.squared.mcfadden")), file='Table-S3.docx')Table S4: VOIs, Control regions and stock metrics:
formula=paste("Rel_p1Y_pos~log_CMC+prof.Bilat_NAcc_mask+prof.Bilat_MPFC_mask+prof.Bilat_AIns_mask+prof.Left_PIns_mask+prof.Left_PVis_mask",sep="")
print(formula)[1] "Rel_p1Y_pos~log_CMC+prof.Bilat_NAcc_mask+prof.Bilat_MPFC_mask+prof.Bilat_AIns_mask+prof.Left_PIns_mask+prof.Left_PVis_mask"out_prof_control3 <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_prof_control3,digits = getOption("jtools-digits", default = 3),scale=TRUE))MODEL INFO:
Observations: 44
Dependent Variable: Rel_p1Y_pos
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(6) = 12.484, p = 0.052
Pseudo-R² (Cragg-Uhler) = 0.329
Pseudo-R² (McFadden) = 0.205
AIC = 62.513, BIC = 75.002
Standard errors: MLE
------------------------------------------------------------
Est. S.E. z val. p
-------------------------- -------- ------- -------- -------
(Intercept) 0.026 0.353 0.074 0.941
log_CMC 0.190 0.372 0.510 0.610
prof.Bilat_NAcc_mask 1.601 0.686 2.334 0.020
prof.Bilat_MPFC_mask 0.265 0.611 0.433 0.665
prof.Bilat_AIns_mask -0.630 0.588 -1.071 0.284
prof.Left_PIns_mask -0.017 0.461 -0.038 0.970
prof.Left_PVis_mask -0.567 0.497 -1.139 0.255
------------------------------------------------------------
Continuous predictors are mean-centered and scaled by 1 s.d. The outcome variable remains in its original units.formula=paste("Rel_p1Y_pos~Rel_m3Y_pos+graph.Bilat_NAcc_mask+graph.Bilat_MPFC_mask+graph.Bilat_AIns_mask+graph.Left_PIns_mask+graph.Left_PVis_mask",sep="")
print(formula)[1] "Rel_p1Y_pos~Rel_m3Y_pos+graph.Bilat_NAcc_mask+graph.Bilat_MPFC_mask+graph.Bilat_AIns_mask+graph.Left_PIns_mask+graph.Left_PVis_mask"out_graph_control3 <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_graph_control3,digits = getOption("jtools-digits", default = 3),scale=TRUE))MODEL INFO:
Observations: 44
Dependent Variable: Rel_p1Y_pos
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(6) = 13.089, p = 0.042
Pseudo-R² (Cragg-Uhler) = 0.343
Pseudo-R² (McFadden) = 0.215
AIC = 61.908, BIC = 74.398
Standard errors: MLE
-------------------------------------------------------------
Est. S.E. z val. p
--------------------------- -------- ------- -------- -------
(Intercept) -0.679 0.642 -1.057 0.291
Rel_m3Y_pos 1.069 0.855 1.250 0.211
graph.Bilat_NAcc_mask 2.153 0.900 2.391 0.017
graph.Bilat_MPFC_mask -0.757 0.599 -1.265 0.206
graph.Bilat_AIns_mask -0.342 0.762 -0.449 0.653
graph.Left_PIns_mask -0.275 0.472 -0.583 0.560
graph.Left_PVis_mask 0.225 0.530 0.424 0.672
-------------------------------------------------------------
Continuous predictors are mean-centered and scaled by 1 s.d. The outcome variable remains in its original units.formula=paste("Rel_p1Y_pos~f_s+f_ebit+f_ebitmargin+f_roe+f_roic+fund.Bilat_NAcc_mask+fund.Bilat_MPFC_mask+fund.Bilat_AIns_mask+fund.Left_PIns_mask+fund.Left_PVis_mask",sep="")
print(formula)[1] "Rel_p1Y_pos~f_s+f_ebit+f_ebitmargin+f_roe+f_roic+fund.Bilat_NAcc_mask+fund.Bilat_MPFC_mask+fund.Bilat_AIns_mask+fund.Left_PIns_mask+fund.Left_PVis_mask"out_fund_control3 <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_fund_control3,digits = getOption("jtools-digits", default = 3),scale=TRUE))MODEL INFO:
Observations: 44
Dependent Variable: Rel_p1Y_pos
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(10) = 14.925, p = 0.135
Pseudo-R² (Cragg-Uhler) = 0.384
Pseudo-R² (McFadden) = 0.245
AIC = 68.072, BIC = 87.698
Standard errors: MLE
------------------------------------------------------------
Est. S.E. z val. p
-------------------------- -------- ------- -------- -------
(Intercept) 0.024 0.370 0.066 0.947
f_s -0.294 0.506 -0.581 0.561
f_ebit 0.179 0.451 0.398 0.690
f_ebitmargin 0.767 0.600 1.278 0.201
f_roe -0.604 0.872 -0.693 0.488
f_roic 0.103 0.634 0.163 0.871
fund.Bilat_NAcc_mask 2.183 0.946 2.308 0.021
fund.Bilat_MPFC_mask 0.393 0.597 0.659 0.510
fund.Bilat_AIns_mask -1.651 0.926 -1.782 0.075
fund.Left_PIns_mask -1.620 0.669 -2.420 0.016
fund.Left_PVis_mask 0.893 0.549 1.626 0.104
------------------------------------------------------------
Continuous predictors are mean-centered and scaled by 1 s.d. The outcome variable remains in its original units.formula=paste("Rel_p1Y_pos~rv_evs+rv_ebitda+rv_pe+rv_pb+rela.Bilat_NAcc_mask+rela.Bilat_MPFC_mask+rela.Bilat_AIns_mask+rela.Left_PIns_mask+rela.Left_PVis_mask",sep="")
print(formula)[1] "Rel_p1Y_pos~rv_evs+rv_ebitda+rv_pe+rv_pb+rela.Bilat_NAcc_mask+rela.Bilat_MPFC_mask+rela.Bilat_AIns_mask+rela.Left_PIns_mask+rela.Left_PVis_mask"out_rela_control3 <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_rela_control3,digits = getOption("jtools-digits", default = 3),scale=TRUE))MODEL INFO:
Observations: 44
Dependent Variable: Rel_p1Y_pos
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(9) = 10.149, p = 0.339
Pseudo-R² (Cragg-Uhler) = 0.275
Pseudo-R² (McFadden) = 0.166
AIC = 70.848, BIC = 88.690
Standard errors: MLE
------------------------------------------------------------
Est. S.E. z val. p
-------------------------- -------- ------- -------- -------
(Intercept) 0.102 0.437 0.233 0.816
rv_evs 0.262 0.647 0.406 0.685
rv_ebitda 1.199 0.747 1.605 0.108
rv_pe -0.729 0.699 -1.043 0.297
rv_pb 0.703 1.769 0.397 0.691
rela.Bilat_NAcc_mask 0.865 0.613 1.410 0.159
rela.Bilat_MPFC_mask 0.832 0.610 1.363 0.173
rela.Bilat_AIns_mask -0.852 0.588 -1.448 0.148
rela.Left_PIns_mask -0.799 0.632 -1.265 0.206
rela.Left_PVis_mask 0.069 0.494 0.140 0.888
------------------------------------------------------------
Continuous predictors are mean-centered and scaled by 1 s.d. The outcome variable remains in its original units.formula=paste("Rel_p1Y_pos~news_incgr+news.Bilat_NAcc_mask+news.Bilat_MPFC_mask+news.Bilat_AIns_mask+news.Left_PIns_mask+news.Left_PVis_mask",sep="")
print(formula)[1] "Rel_p1Y_pos~news_incgr+news.Bilat_NAcc_mask+news.Bilat_MPFC_mask+news.Bilat_AIns_mask+news.Left_PIns_mask+news.Left_PVis_mask"out_news_control3 <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_news_control3,digits = getOption("jtools-digits", default = 3),scale=TRUE))MODEL INFO:
Observations: 44
Dependent Variable: Rel_p1Y_pos
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(6) = 4.569, p = 0.600
Pseudo-R² (Cragg-Uhler) = 0.131
Pseudo-R² (McFadden) = 0.075
AIC = 70.428, BIC = 82.918
Standard errors: MLE
------------------------------------------------------------
Est. S.E. z val. p
-------------------------- -------- ------- -------- -------
(Intercept) 0.159 0.400 0.399 0.690
news_incgr 1.241 1.523 0.815 0.415
news.Bilat_NAcc_mask -0.280 0.505 -0.555 0.579
news.Bilat_MPFC_mask 0.359 0.531 0.677 0.498
news.Bilat_AIns_mask 0.449 0.505 0.888 0.375
news.Left_PIns_mask -0.297 0.442 -0.671 0.502
news.Left_PVis_mask 0.002 0.375 0.006 0.995
------------------------------------------------------------
Continuous predictors are mean-centered and scaled by 1 s.d. The outcome variable remains in its original units.coef_names <- c("Current Market Capitalization" = "log_CMC","Profile - PVis" = "prof.Left_PVis_mask","Profile - PIns" = "prof.Left_PIns_mask","Profile - NAcc" = "prof.Bilat_NAcc_mask", "Profile - MPFC" = "prof.Bilat_MPFC_mask", "Profile - AIns" = "prof.Bilat_AIns_mask",
"36M Performance"='Rel_m3Y_pos',"Graph - PVis" = "graph.Left_PVis_mask","Graph - PIns" = "graph.Left_PIns_mask","Graph - NAcc" = "graph.Bilat_NAcc_mask", "Graph - MPFC" = "graph.Bilat_MPFC_mask", "Graph - AIns" = "graph.Bilat_AIns_mask",
"Sales" ="f_s", "EBIT" = "f_ebit", "EBIT Margin (%)" = "f_ebitmargin","ROE (%)" = "f_roe","ROIC (%)" = "f_roic",
"Fundamentals - PVis" = "fund.Left_PVis_mask","Fundamentals - PIns" = "fund.Left_PIns_mask","Fundamentals - NAcc" = "fund.Bilat_NAcc_mask", "Fundamentals - MPFC" = "fund.Bilat_MPFC_mask", "Fundamentals - AIns" = "fund.Bilat_AIns_mask",
"EV/Sales" = "rv_evs","EV/EBITDA" = "rv_ebitda","P/E" = "rv_pe","P/B" = "rv_pb",
"Relative Valuation - PVis" = "rela.Left_PVis_mask","Relative Valuation - PIns" = "rela.Left_PIns_mask","Relative Valuation - NAcc" = "rela.Bilat_NAcc_mask", "Relative Valuation - MPFC" = "rela.Bilat_MPFC_mask", "Relative Valuation - AIns" = "rela.Bilat_AIns_mask",
"Income growth" = "news_incgr", "News item - PVis" = "news.Left_PVis_mask","News item - PIns" = "news.Left_PIns_mask","News item - NAcc" = "news.Bilat_NAcc_mask", "News item - MPFC" = "news.Bilat_MPFC_mask", "News item - AIns" = "news.Bilat_AIns_mask"
)
# Uncomment to create table
#quick_docx(export_summs(out_prof_control3,out_graph_control3,out_fund_control3,out_rela_control3,out_news_control3, scale = TRUE,coefs = coef_names,error_pos = c("same"),
# model.names=c("Company Profile","Price Graph","Fundamentals","Relative Valuation","News Item"),
# statistics= c(N = "nobs", AIC = "AIC", "Pseudo R2 (McFadden)" = "pseudo.r.squared.mcfadden")), file='Table-S4.docx')Testing if there is a true relationship between PIns activity at fundamentals screen and future stock performance:
formula=paste("Rel_p1Y_pos~fund.Left_PIns_mask",sep="")
print(formula)[1] "Rel_p1Y_pos~fund.Left_PIns_mask"out_fund_pins <- glm(formula,data=reg_df,family = binomial(link = logit))
print(summ(out_fund_pins,digits = getOption("jtools-digits", default = 3),scale=TRUE))MODEL INFO:
Observations: 44
Dependent Variable: Rel_p1Y_pos
Type: Generalized linear model
Family: binomial
Link function: logit
MODEL FIT:
χ²(1) = 0.535, p = 0.464
Pseudo-R² (Cragg-Uhler) = 0.016
Pseudo-R² (McFadden) = 0.009
AIC = 64.462, BIC = 68.030
Standard errors: MLE
-----------------------------------------------------------
Est. S.E. z val. p
------------------------- -------- ------- -------- -------
(Intercept) 0.000 0.303 0.000 1.000
fund.Left_PIns_mask -0.225 0.310 -0.726 0.468
-----------------------------------------------------------
Continuous predictors are mean-centered and scaled by 1 s.d. The outcome variable remains in its original units.Compare models:
anova(out_prof, out_prof_control3, test = "LRT")Analysis of Deviance Table
Model 1: Rel_p1Y_pos ~ log_CMC + prof.Bilat_NAcc_mask + prof.Bilat_MPFC_mask +
prof.Bilat_AIns_mask
Model 2: Rel_p1Y_pos ~ log_CMC + prof.Bilat_NAcc_mask + prof.Bilat_MPFC_mask +
prof.Bilat_AIns_mask + prof.Left_PIns_mask + prof.Left_PVis_mask
Resid. Df Resid. Dev Df Deviance Pr(>Chi)
1 39 50.454
2 37 48.513 2 1.9408 0.3789anova(out_graph, out_graph_control3, test = "LRT")Analysis of Deviance Table
Model 1: Rel_p1Y_pos ~ Rel_m3Y_pos + graph.Bilat_NAcc_mask + graph.Bilat_MPFC_mask +
graph.Bilat_AIns_mask
Model 2: Rel_p1Y_pos ~ Rel_m3Y_pos + graph.Bilat_NAcc_mask + graph.Bilat_MPFC_mask +
graph.Bilat_AIns_mask + graph.Left_PIns_mask + graph.Left_PVis_mask
Resid. Df Resid. Dev Df Deviance Pr(>Chi)
1 39 48.281
2 37 47.908 2 0.37255 0.83anova(out_fund, out_fund_control3, test = "LRT")Analysis of Deviance Table
Model 1: Rel_p1Y_pos ~ f_s + f_ebit + f_ebitmargin + f_roe + f_roic +
fund.Bilat_NAcc_mask + fund.Bilat_MPFC_mask + fund.Bilat_AIns_mask
Model 2: Rel_p1Y_pos ~ f_s + f_ebit + f_ebitmargin + f_roe + f_roic +
fund.Bilat_NAcc_mask + fund.Bilat_MPFC_mask + fund.Bilat_AIns_mask +
fund.Left_PIns_mask + fund.Left_PVis_mask
Resid. Df Resid. Dev Df Deviance Pr(>Chi)
1 35 54.013
2 33 46.072 2 7.9408 0.01887 *
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1anova(out_rela, out_rela_control3, test = "LRT")Analysis of Deviance Table
Model 1: Rel_p1Y_pos ~ rv_evs + rv_ebitda + rv_pe + rv_pb + rela.Bilat_NAcc_mask +
rela.Bilat_MPFC_mask + rela.Bilat_AIns_mask
Model 2: Rel_p1Y_pos ~ rv_evs + rv_ebitda + rv_pe + rv_pb + rela.Bilat_NAcc_mask +
rela.Bilat_MPFC_mask + rela.Bilat_AIns_mask + rela.Left_PIns_mask +
rela.Left_PVis_mask
Resid. Df Resid. Dev Df Deviance Pr(>Chi)
1 36 52.646
2 34 50.848 2 1.7978 0.407anova(out_news, out_news_control3, test = "LRT")Analysis of Deviance Table
Model 1: Rel_p1Y_pos ~ news_incgr + news.Bilat_NAcc_mask + news.Bilat_MPFC_mask +
news.Bilat_AIns_mask
Model 2: Rel_p1Y_pos ~ news_incgr + news.Bilat_NAcc_mask + news.Bilat_MPFC_mask +
news.Bilat_AIns_mask + news.Left_PIns_mask + news.Left_PVis_mask
Resid. Df Resid. Dev Df Deviance Pr(>Chi)
1 39 56.896
2 37 56.428 2 0.46731 0.7916