---------------------------------------------------------------------------------------------------------------------------------------- name: log: C:\Users\56570tpe\Dropbox\work\transfer values\stata\replication files\allgamesindividuals.log log type: text opened on: 7 Aug 2017, 15:52:46 . . use "dataset players.dta" . . *****************************************************VARIABLE GENERATION************************************************** . . *generate the team values for Transfermarkt . bysort matchteam_id: egen tot=total(value) . bysort matchteam_id: egen avvalue=mean(value) . bysort matchteam_id: egen sd=sd(value) . bysort matchteam_id: egen max=max(value) . bysort matchteam_id: egen min=min(value) . bysort matchteam_id: egen med=median(value) . bysort matchteam_id: egen number=count(value) . . *generate values for well supported teams in the crowd . gen playteamsupported=1 if playteamrelfans>0.01 (36,226 missing values generated) . replace playteamsupported=0 if playteamsupported==. (36,226 real changes made) . . bysort matchteam_id: egen numbersupported=total(playteamsupported) . . gen indsupportvalue=playteamrelfans*value (1,206 missing values generated) . bysort matchteam_id: egen support=mean(playteamfans+1) (1206 missing values generated) . bysort matchteam_id: egen supportvalue=mean(indsupportvalue) (1206 missing values generated) . . *************************************************************rematch at the game level**************************************** . . collapse day month year home homereal uefa conmebol fifapoints fifarank wcqual10 wcqual14 euro12qual euro12 europo copam wc10 match_id > team_id /// > res go margin tot* av* sd max min med number* support supportvalue playteamrel* winodds drawodds /// > lossodds odd*, by(matchteam_id team group) . . sort match_id home . . tempfile playerlevel . . save `playerlevel' file C:\Users\56570tpe\AppData\Local\Temp\ST_0k000001.tmp saved . . for var home homereal team_id uefa conmebol fifapoints fifarank go res margin matchteam_id team tot av* sd max min med number /// > support supportvalue playteamrel* numbersupported winodds: rename X op_X -> rename home op_home -> rename homereal op_homereal -> rename team_id op_team_id -> rename uefa op_uefa -> rename conmebol op_conmebol -> rename fifapoints op_fifapoints -> rename fifarank op_fifarank -> rename go op_go -> rename res op_res -> rename margin op_margin -> rename matchteam_id op_matchteam_id -> rename team op_team -> rename tot op_tot -> rename avvalue op_avvalue -> rename sd op_sd -> rename max op_max -> rename min op_min -> rename med op_med -> rename number op_number -> rename support op_support -> rename supportvalue op_supportvalue -> rename playteamrelfans op_playteamrelfans -> rename playteamrelcountryfans op_playteamrelcountryfans -> rename numbersupported op_numbersupported -> rename winodds op_winodds . rename lossodds winodds . rename op_winodds lossodds . . gen home=abs(op_home-1) . . sort match_id home . . merge 1:1 match_id home using `playerlevel', update replace Result # of obs. ----------------------------------------- not matched 0 matched 2,134 not updated 0 (_merge==3) missing updated 0 (_merge==4) nonmissing conflict 2,134 (_merge==5) ----------------------------------------- . . drop _merge . drop if avvalue==. (0 observations deleted) . drop if op_avvalue==. (0 observations deleted) . gen date=mdy(month,day,year) . . tab uefa conmebol (mean) | (mean) conmebol uefa | 0 1 | Total -----------+----------------------+---------- 0 | 54 396 | 450 1 | 1,684 0 | 1,684 -----------+----------------------+---------- Total | 1,738 396 | 2,134 . drop if conmebol==uefa (54 observations deleted) . drop if op_conmebol==op_uefa (40 observations deleted) . . ************************Make variables to merge the ELO dataset**************** . sort date . gen seqnum=_n . bysort team_id op_team_id year month: egen lastmonthgame=max(seqnum) . bysort team_id op_team_id year month: egen firstmonthgame=min(seqnum) . . gen monthgamenumber=2 if lastmonthgame==seqnum (46 missing values generated) . replace monthgamenumber=1 if firstmonthgame==seqnum (1,994 real changes made) . . sum month* Variable | Obs Mean Std. Dev. Min Max -------------+--------------------------------------------------------- month | 2,040 8.151961 2.385884 2 11 monthgamen~r | 2,040 1.022549 .148497 1 2 . . save allgamesplayers.dta, replace file allgamesplayers.dta saved . . *******************************************ADD THE ELO DATA*************************************** . . use elofile.dta, replace . . ************************Make variables to merge**************** . . gen date=mdy(month,day,year) (14 missing values generated) . sort date . gen seqnum=_n . bysort team_id op_team_id year month: egen lastmonthgame=max(seqnum) . bysort team_id op_team_id year month: egen firstmonthgame=min(seqnum) . . gen monthgamenumber=2 if lastmonthgame==seqnum (46 missing values generated) . replace monthgamenumber=1 if firstmonthgame==seqnum (1,994 real changes made) . . sum month* Variable | Obs Mean Std. Dev. Min Max -------------+--------------------------------------------------------- month | 2,040 8.151961 2.385884 2 11 monthgamen~r | 2,040 1.022549 .148497 1 2 . . tempfile elofile2 . . save `elofile2' file C:\Users\56570tpe\AppData\Local\Temp\ST_0k000002.tmp saved . . use allgamesplayers.dta, replace . . merge 1:1 team_id op_team_id year month monthgamenumber using `elofile2' (note: variable team was str20, now str22 to accommodate using data's values) (note: variable op_team was str20, now str22 to accommodate using data's values) Result # of obs. ----------------------------------------- not matched 0 matched 2,040 (_merge==3) ----------------------------------------- . . ***Check the correpondance . . drop if matchteam_id==. (0 observations deleted) . drop check . . sort match_id . gen check=1 if elostart[_n]==op_elostart[_n-1] & match_id[_n]==match_id[_n-1] (1,020 missing values generated) . replace check=1 if elostart[_n]==op_elostart[_n+1] & match_id[_n]==match_id[_n+1] (1,020 real changes made) . sum elostart op_elostart date check Variable | Obs Mean Std. Dev. Min Max -------------+--------------------------------------------------------- elostart | 2,040 1643.373 247.4678 857 2140 op_elostart | 2,040 1643.373 247.4678 857 2140 date | 2,040 18696.77 641.0821 17452 19681 check | 2,040 1 0 1 1 . . *****************************SET UP VARIABLE FOR RANDOM SELECTION OF 1 OBS PER GAME *********************** . *each 1st game home, second away . sort match_id home . gen selectorsetup=_n-int(_n/4)*4 . tab selectorsetup selectorset | up | Freq. Percent Cum. ------------+----------------------------------- 0 | 510 25.00 25.00 1 | 510 25.00 50.00 2 | 510 25.00 75.00 3 | 510 25.00 100.00 ------------+----------------------------------- Total | 2,040 100.00 . . gen selector=1 if selectorsetup==0 (1,530 missing values generated) . replace selector=1 if selectorsetup==1 (510 real changes made) . replace selector=0 if selector==. (1,020 real changes made) . . bysort match_id: egen meanselector=mean(selector) . sum meanselector Variable | Obs Mean Std. Dev. Min Max -------------+--------------------------------------------------------- meanselector | 2,040 .5 0 .5 .5 . . tab homereal selector (mean) | selector homereal | 0 1 | Total -----------+----------------------+---------- 0 | 545 543 | 1,088 1 | 475 477 | 952 -----------+----------------------+---------- Total | 1,020 1,020 | 2,040 . . *set constraint for oprobit . constraint def 1 [cut1]_cons = -[cut2]_cons . . ***************************************RESCALING AND EXTRA GENERATION********************************* . *generate the wins-points . gen win=1 if res==2 (1,245 missing values generated) . replace win=0 if win==. (1,245 real changes made) . gen draw=1 if res==1 (1,590 missing values generated) . replace draw=0 if draw==. (1,590 real changes made) . gen loss=1 if res==0 (1,245 missing values generated) . replace loss=0 if loss==. (1,245 real changes made) . . **For estimation on the difference . gen difhomereal=homereal-op_homereal . gen ldifavvalue=log(avvalue/op_avvalue) . gen ldifsq=log(avvalue/op_avvalue)^2 . gen ldifsd=log(sd/avvalue*op_avvalue/op_sd) (14 missing values generated) . gen ldifmax=log(max/avvalue*op_avvalue/op_max) . gen ldifmin=log(min/avvalue*op_avvalue/op_min) . gen ldifmed=log(med/avvalue*op_avvalue/op_med) . . **Alternative forecasters . gen ldifnumber=log(number/op_number) . gen ldiffifapoints=log((fifapoints+1)/(op_fifapoints+1)) . gen ldifelopoints=log(elostart/op_elostart) . . *bias in supported teams? . gen ldifnumbersupported=log((numbersupported+1)/(op_numbersupported+1)) . gen ldifsupport=log(support/op_support) . gen ldifsupportvalue=log((supportvalue+1)/(op_supportvalue+1)) . gen difplayteamrelfans=playteamrelfans-op_playteamrelfans . . egen country_id=group(team_id) . sum country_id team_id uefa conmebol op_uefa op_conmebol Variable | Obs Mean Std. Dev. Min Max -------------+--------------------------------------------------------- country_id | 2,040 32.16863 18.23478 1 63 team_id | 2,040 40.31176 22.8506 1 78 uefa | 2,040 .8147059 .3886312 0 1 conmebol | 2,040 .1852941 .3886312 0 1 op_uefa | 2,040 .8147059 .3886312 0 1 -------------+--------------------------------------------------------- op_conmebol | 2,040 .1852941 .3886312 0 1 . . save allgamesplayers.dta, replace file allgamesplayers.dta saved . . . *************************************************************ESTIMATION MATCH OUTCOME PART******************************************** > ************************** . local games=r(N)/2 . . **summary stats of estimation sample => Table 1 and 2 in paper . sum loss draw win margin lossodds drawodds winodds odd_win odd_draw odd_loss avvalue op_avvalue fifapoints op_fifapoints elostart op_e > lostart /// > homereal op_homereal number* op_number* support* op_support* playteamrelfans op_playteamrelfans wcqual10 wcqual14 euro12 euro12qual co > pa* europo wc10 if selector==1 Variable | Obs Mean Std. Dev. Min Max -------------+--------------------------------------------------------- loss | 1,020 .3686275 .4826693 0 1 draw | 1,020 .2205882 .4148467 0 1 win | 1,020 .4107843 .4922175 0 1 margin | 1,020 .1029412 2.230177 -10 11 lossodds | 1,020 .3881211 .2624673 .0133012 .9452337 -------------+--------------------------------------------------------- drawodds | 1,020 .2222055 .0746592 .0423176 .5399877 winodds | 1,020 .3896734 .2628513 .0124486 .9396976 odd_win | 1,020 6.260078 10.04058 1.01 76.69 odd_draw | 1,020 5.180039 3.40923 1.69 22.56 odd_loss | 1,020 6.151225 9.550688 1.01 71.35 -------------+--------------------------------------------------------- avvalue | 1,020 4693453 5407573 1 2.86e+07 op_avvalue | 1,020 4821070 5839096 1 3.18e+07 fifapoints | 1,020 688.1539 350.2975 0 1883 op_fifapoi~s | 1,020 699.2382 366.6184 0 1883 elostart | 1,020 1641.455 247.3327 857 2140 -------------+--------------------------------------------------------- op_elostart | 1,020 1645.291 247.7094 859 2136 homereal | 1,020 .4676471 .499197 0 1 op_homereal | 1,020 .4656863 .4990659 0 1 number | 1,020 19.62647 2.200608 16 23 numbersupp~d | 1,020 2.307843 3.722638 0 23 -------------+--------------------------------------------------------- op_number | 1,020 19.61667 2.221122 15 23 op_numbers~d | 1,020 2.133333 3.160084 0 22 support | 1,020 254.3855 504.5734 1 3463.428 supportvalue | 1,020 71528.23 203088.4 0 2035881 op_support | 1,020 209.1215 390.3092 1 3514.444 -------------+--------------------------------------------------------- op_support~e | 1,020 57751.48 165419.5 0 1815932 playte~lfans | 1,020 .0053416 .01064 0 .0730095 op_pla~lfans | 1,020 .0043867 .0082307 0 .0740889 wcqual10 | 1,020 .2627451 .440341 0 1 wcqual14 | 1,020 .4215686 .4940524 0 1 -------------+--------------------------------------------------------- euro12 | 1,020 .0303922 .171748 0 1 euro12qual | 1,020 .2352941 .4243906 0 1 copam | 1,020 .0196078 .1387164 0 1 europo | 1,020 .0078431 .0882568 0 1 wc10 | 1,020 .022549 .1485335 0 1 . . ************************************BASE MODEL******************************************* . . *Baseline regression . oprobit res difhomereal ldifavvalue ldifnumber if selector==1, constraint(1) Iteration 0: log likelihood = -1089.2568 Iteration 1: log likelihood = -825.66491 Iteration 2: log likelihood = -820.19207 Iteration 3: log likelihood = -820.17789 Iteration 4: log likelihood = -820.17789 Ordered probit regression Number of obs = 1,020 Wald chi2(3) = 377.69 Log likelihood = -820.17789 Prob > chi2 = 0.0000 ( 1) [cut1]_cons + [cut2]_cons = 0 ------------------------------------------------------------------------------ res | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- difhomereal | .2806607 .042243 6.64 0.000 .197866 .3634554 ldifavvalue | .4504088 .0243791 18.48 0.000 .4026266 .498191 ldifnumber | 1.775337 .8797737 2.02 0.044 .0510126 3.499662 -------------+---------------------------------------------------------------- /cut1 | -.4088783 .0242953 -.4564963 -.3612604 /cut2 | .4088783 .0242953 .3612604 .4564963 ------------------------------------------------------------------------------ Note: 7 observations completely determined. Standard errors questionable. . estat ic, n(`games') Akaike's information criterion and Bayesian information criterion ----------------------------------------------------------------------------- Model | Obs ll(null) ll(model) df AIC BIC -------------+--------------------------------------------------------------- . | 1,020 . -820.1779 4 1648.356 1668.066 ----------------------------------------------------------------------------- Note: N=1020 used in calculating BIC. . predict losshat drawhat winhat (option pr assumed; predicted probabilities) . outreg2 using Table3a.xls, replace keep(difhomereal ldifavvalue ldifnumber) addstat(Log Likelihood, e(ll)) slow(750) dec(3) ctitle(Cro > wd) Table3a.xls dir : seeout . outreg2 using Table8.xls, replace keep(difhomereal ldifavvalue ldifnumber) addstat(Log Likelihood, e(ll)) slow(750) dec(3) ctitle(Game > result) Table8.xls dir : seeout . outreg2 using Tableapp.xls, replace keep(difhomereal ldifavvalue ldifnumber) addstat(Log Likelihood, e(ll)) slow(750) dec(3) ctitle(ba > se) Tableapp.xls dir : seeout . outreg2 using Table3b.xls, replace keep(difhomereal ldifavvalue ldifnumber) addstat(Log Likelihood, e(ll)) slow(750) dec(3) ctitle(opr > obit) Table3b.xls dir : seeout . . *other tries with distributional measures of TM values: . oprobit res difhomereal ldifavvalue ldifnumber ldifsd if selector==1, constraint(1) Iteration 0: log likelihood = -1082.6535 Iteration 1: log likelihood = -819.5217 Iteration 2: log likelihood = -818.58496 Iteration 3: log likelihood = -818.58413 Iteration 4: log likelihood = -818.58413 Ordered probit regression Number of obs = 1,013 Wald chi2(4) = 378.13 Log likelihood = -818.58413 Prob > chi2 = 0.0000 ( 1) [cut1]_cons + [cut2]_cons = 0 ------------------------------------------------------------------------------ res | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- difhomereal | .2815127 .0422772 6.66 0.000 .1986509 .3643745 ldifavvalue | .4305785 .0267725 16.08 0.000 .3781054 .4830516 ldifnumber | 1.821137 .8781973 2.07 0.038 .0999019 3.542372 ldifsd | -.1401467 .0785109 -1.79 0.074 -.2940253 .0137319 -------------+---------------------------------------------------------------- /cut1 | -.4099897 .0243619 -.4577381 -.3622414 /cut2 | .4099897 .0243619 .3622414 .4577381 ------------------------------------------------------------------------------ . estat ic, n(`games') Akaike's information criterion and Bayesian information criterion ----------------------------------------------------------------------------- Model | Obs ll(null) ll(model) df AIC BIC -------------+--------------------------------------------------------------- . | 1,020 . -818.5841 5 1647.168 1671.806 ----------------------------------------------------------------------------- Note: N=1020 used in calculating BIC. . outreg2 using Tableapp.xls, append keep(difhomereal ldifavvalue ldifnumber ldifsd) addstat(Log Likelihood, e(ll)) slow(750) dec(3) ct > itle(sd) Tableapp.xls dir : seeout . . oprobit res difhomereal ldifavvalue ldifnumber ldifmax if selector==1, constraint(1) Iteration 0: log likelihood = -1089.2568 Iteration 1: log likelihood = -825.60644 Iteration 2: log likelihood = -818.66031 Iteration 3: log likelihood = -818.64181 Iteration 4: log likelihood = -818.64181 Ordered probit regression Number of obs = 1,020 Wald chi2(4) = 378.19 Log likelihood = -818.64181 Prob > chi2 = 0.0000 ( 1) [cut1]_cons + [cut2]_cons = 0 ------------------------------------------------------------------------------ res | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- difhomereal | .2807727 .0422726 6.64 0.000 .1979199 .3636256 ldifavvalue | .4345261 .0260095 16.71 0.000 .3835485 .4855037 ldifnumber | 1.83803 .8787802 2.09 0.036 .1156528 3.560408 ldifmax | -.1210763 .0690886 -1.75 0.080 -.2564875 .0143348 -------------+---------------------------------------------------------------- /cut1 | -.4099522 .0243596 -.4576962 -.3622082 /cut2 | .4099522 .0243596 .3622082 .4576962 ------------------------------------------------------------------------------ Note: 7 observations completely determined. Standard errors questionable. . estat ic, n(`games') Akaike's information criterion and Bayesian information criterion ----------------------------------------------------------------------------- Model | Obs ll(null) ll(model) df AIC BIC -------------+--------------------------------------------------------------- . | 1,020 . -818.6418 5 1647.284 1671.921 ----------------------------------------------------------------------------- Note: N=1020 used in calculating BIC. . outreg2 using Tableapp.xls, append keep(difhomereal ldifavvalue ldifnumber ldifmax) addstat(Log Likelihood, e(ll)) slow(750) dec(3) c > title(max) Tableapp.xls dir : seeout . . oprobit res difhomereal ldifavvalue ldifnumber ldifmin if selector==1, constraint(1) Iteration 0: log likelihood = -1089.2568 Iteration 1: log likelihood = -829.86521 Iteration 2: log likelihood = -820.15379 Iteration 3: log likelihood = -820.11145 Iteration 4: log likelihood = -820.11145 Ordered probit regression Number of obs = 1,020 Wald chi2(4) = 377.20 Log likelihood = -820.11145 Prob > chi2 = 0.0000 ( 1) [cut1]_cons + [cut2]_cons = 0 ------------------------------------------------------------------------------ res | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- difhomereal | .2811162 .042267 6.65 0.000 .1982745 .363958 ldifavvalue | .4558521 .028635 15.92 0.000 .3997285 .5119757 ldifnumber | 1.77148 .8794004 2.01 0.044 .0478866 3.495073 ldifmin | -.0026082 .0071556 -0.36 0.715 -.0166329 .0114166 -------------+---------------------------------------------------------------- /cut1 | -.408813 .024291 -.4564224 -.3612035 /cut2 | .408813 .024291 .3612035 .4564224 ------------------------------------------------------------------------------ Note: 7 observations completely determined. Standard errors questionable. . estat ic, n(`games') Akaike's information criterion and Bayesian information criterion ----------------------------------------------------------------------------- Model | Obs ll(null) ll(model) df AIC BIC -------------+--------------------------------------------------------------- . | 1,020 . -820.1114 5 1650.223 1674.861 ----------------------------------------------------------------------------- Note: N=1020 used in calculating BIC. . outreg2 using Tableapp.xls, append keep(difhomereal ldifavvalue ldifnumber ldifmin) addstat(Log Likelihood, e(ll)) slow(750) dec(3) c > title(min) Tableapp.xls dir : seeout . . oprobit res difhomereal ldifavvalue ldifnumber ldifmed if selector==1, constraint(1) Iteration 0: log likelihood = -1089.2568 Iteration 1: log likelihood = -826.08646 Iteration 2: log likelihood = -820.1653 Iteration 3: log likelihood = -820.14342 Iteration 4: log likelihood = -820.14342 Ordered probit regression Number of obs = 1,020 Wald chi2(4) = 376.92 Log likelihood = -820.14342 Prob > chi2 = 0.0000 ( 1) [cut1]_cons + [cut2]_cons = 0 ------------------------------------------------------------------------------ res | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- difhomereal | .2810948 .0422822 6.65 0.000 .1982232 .3639665 ldifavvalue | .4477022 .0264696 16.91 0.000 .3958228 .4995816 ldifnumber | 1.7775 .8795467 2.02 0.043 .0536204 3.50138 ldifmed | .0040041 .0152765 0.26 0.793 -.0259373 .0339456 -------------+---------------------------------------------------------------- /cut1 | -.4088376 .0242929 -.4564508 -.3612244 /cut2 | .4088376 .0242929 .3612244 .4564508 ------------------------------------------------------------------------------ Note: 7 observations completely determined. Standard errors questionable. . estat ic, n(`games') Akaike's information criterion and Bayesian information criterion ----------------------------------------------------------------------------- Model | Obs ll(null) ll(model) df AIC BIC -------------+--------------------------------------------------------------- . | 1,020 . -820.1434 5 1650.287 1674.925 ----------------------------------------------------------------------------- Note: N=1020 used in calculating BIC. . outreg2 using Tableapp.xls, append keep(difhomereal ldifavvalue ldifnumber ldifmed) addstat(Log Likelihood, e(ll)) slow(750) dec(3) c > title(median) Tableapp.xls dir : seeout . . oprobit res difhomereal ldifavvalue ldifnumber ldifsq if selector==1, constraint(1) Iteration 0: log likelihood = -1089.2568 Iteration 1: log likelihood = -825.2568 Iteration 2: log likelihood = -819.57056 Iteration 3: log likelihood = -819.54566 Iteration 4: log likelihood = -819.54566 Ordered probit regression Number of obs = 1,020 Wald chi2(4) = 376.96 Log likelihood = -819.54566 Prob > chi2 = 0.0000 ( 1) [cut1]_cons + [cut2]_cons = 0 ------------------------------------------------------------------------------ res | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- difhomereal | .2813038 .042237 6.66 0.000 .1985209 .3640868 ldifavvalue | .4515108 .0244938 18.43 0.000 .4035039 .4995176 ldifnumber | 1.745974 .8800264 1.98 0.047 .0211539 3.470794 ldifsq | .0088449 .0078498 1.13 0.260 -.0065404 .0242303 -------------+---------------------------------------------------------------- /cut1 | -.4093594 .0243257 -.4570369 -.361682 /cut2 | .4093594 .0243257 .361682 .4570369 ------------------------------------------------------------------------------ Note: 3 observations completely determined. Standard errors questionable. . estat ic, n(`games') Akaike's information criterion and Bayesian information criterion ----------------------------------------------------------------------------- Model | Obs ll(null) ll(model) df AIC BIC -------------+--------------------------------------------------------------- . | 1,020 . -819.5457 5 1649.091 1673.729 ----------------------------------------------------------------------------- Note: N=1020 used in calculating BIC. . outreg2 using Tableapp.xls, append keep(difhomereal ldifavvalue ldifnumber ldifsq) addstat(Log Likelihood, e(ll)) slow(750) dec(3) ct > itle(median) Tableapp.xls dir : seeout . . oprobit res difhomereal ldifavvalue ldifnumber ldifsd ldifmax ldifmin ldifmed if selector==1, constraint(1) Iteration 0: log likelihood = -1082.6535 Iteration 1: log likelihood = -820.53623 Iteration 2: log likelihood = -818.21481 Iteration 3: log likelihood = -818.21181 Iteration 4: log likelihood = -818.21181 Ordered probit regression Number of obs = 1,013 Wald chi2(7) = 378.20 Log likelihood = -818.21181 Prob > chi2 = 0.0000 ( 1) [cut1]_cons + [cut2]_cons = 0 ------------------------------------------------------------------------------ res | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- difhomereal | .2819811 .042364 6.66 0.000 .1989492 .365013 ldifavvalue | .4434982 .031001 14.31 0.000 .3827373 .5042591 ldifnumber | 1.819198 .8782719 2.07 0.038 .0978167 3.540579 ldifsd | -.1563197 .2942922 -0.53 0.595 -.7331219 .4204824 ldifmax | -.0071356 .2548071 -0.03 0.978 -.5065484 .4922772 ldifmin | -.0058209 .0074081 -0.79 0.432 -.0203405 .0086986 ldifmed | -.0058432 .0161341 -0.36 0.717 -.0374654 .025779 -------------+---------------------------------------------------------------- /cut1 | -.4101013 .0243685 -.4578627 -.3623399 /cut2 | .4101013 .0243685 .3623399 .4578627 ------------------------------------------------------------------------------ . estat ic, n(`games') Akaike's information criterion and Bayesian information criterion ----------------------------------------------------------------------------- Model | Obs ll(null) ll(model) df AIC BIC -------------+--------------------------------------------------------------- . | 1,020 . -818.2118 8 1652.424 1691.844 ----------------------------------------------------------------------------- Note: N=1020 used in calculating BIC. . outreg2 using Tableapp.xls, append keep(difhomereal ldifavvalue ldifnumber ldifsd ldifmax ldifmin ldifmed) addstat(Log Likelihood, e(l > l)) slow(750) dec(3) ctitle(all) Tableapp.xls dir : seeout . . *tries with other estimation methods: . ologit res difhomereal ldifavvalue ldifnumber if selector==1, constraint(1) Iteration 0: log likelihood = -1089.257 Iteration 1: log likelihood = -825.92218 Iteration 2: log likelihood = -821.68133 Iteration 3: log likelihood = -821.67873 Iteration 4: log likelihood = -821.67873 Ordered logistic regression Number of obs = 1,020 Wald chi2(3) = 318.73 Log likelihood = -821.67873 Prob > chi2 = 0.0000 ( 1) [cut1]_cons + [cut2]_cons = 0 ------------------------------------------------------------------------------ res | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- difhomereal | .4707864 .0720573 6.53 0.000 .3295566 .6120161 ldifavvalue | .7574109 .0443814 17.07 0.000 .6704249 .8443968 ldifnumber | 2.950499 1.522567 1.94 0.053 -.033678 5.934676 -------------+---------------------------------------------------------------- /cut1 | -.6937431 .042536 -.7771122 -.610374 /cut2 | .6937431 .042536 .610374 .7771122 ------------------------------------------------------------------------------ . estat ic, n(`games') Akaike's information criterion and Bayesian information criterion ----------------------------------------------------------------------------- Model | Obs ll(null) ll(model) df AIC BIC -------------+--------------------------------------------------------------- . | 1,020 . -821.6787 4 1651.357 1671.068 ----------------------------------------------------------------------------- Note: N=1020 used in calculating BIC. . predict losslogit drawlogit winlogit (option pr assumed; predicted probabilities) . outreg2 using Table3b.xls, append keep(difhomereal ldifavvalue ldifnumber) addstat(Log Likelihood, e(ll)) slow(750) dec(3) ctitle(olog > it) Table3b.xls dir : seeout . . mlogit res difhomereal ldifavvalue ldifnumber if selector==1, constraint(1) (note: constraint number 1 caused error r(303)) Iteration 0: log likelihood = -1088.093 Iteration 1: log likelihood = -835.503 Iteration 2: log likelihood = -820.76386 Iteration 3: log likelihood = -820.43211 Iteration 4: log likelihood = -820.43195 Iteration 5: log likelihood = -820.43195 Multinomial logistic regression Number of obs = 1,020 LR chi2(6) = 535.32 Prob > chi2 = 0.0000 Log likelihood = -820.43195 Pseudo R2 = 0.2460 ------------------------------------------------------------------------------ res | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- 0 | difhomereal | -.6579443 .1017025 -6.47 0.000 -.8572775 -.458611 ldifavvalue | -1.053536 .0679373 -15.51 0.000 -1.18669 -.9203811 ldifnumber | -4.143334 2.179687 -1.90 0.057 -8.415442 .1287729 _cons | -.1819942 .1005242 -1.81 0.070 -.379018 .0150296 -------------+---------------------------------------------------------------- 1 | difhomereal | -.2113396 .0942016 -2.24 0.025 -.3959713 -.0267079 ldifavvalue | -.476737 .0562277 -8.48 0.000 -.5869412 -.3665328 ldifnumber | -2.773319 1.933971 -1.43 0.152 -6.563832 1.017194 _cons | -.2167216 .0957869 -2.26 0.024 -.4044604 -.0289827 -------------+---------------------------------------------------------------- 2 | (base outcome) ------------------------------------------------------------------------------ . estat ic, n(`games') Akaike's information criterion and Bayesian information criterion ----------------------------------------------------------------------------- Model | Obs ll(null) ll(model) df AIC BIC -------------+--------------------------------------------------------------- . | 1,020 -1088.093 -820.4319 8 1656.864 1696.284 ----------------------------------------------------------------------------- Note: N=1020 used in calculating BIC. . predict lossmlogit drawmlogit winmlogit (option pr assumed; predicted probabilities) . outreg2 using Table3b.xls, append keep(difhomereal ldifavvalue ldifnumber) addstat(Log Likelihood, e(ll)) slow(750) dec(3) ctitle(mlog > it) Table3b.xls dir : seeout . . ********************************************ALTERNATIVE PREDICTORS************************************** . *Odds no estimation needed . *FIFA . oprobit res difhomereal ldiffifapoints ldifnumber if selector==1, constraint(1) Iteration 0: log likelihood = -1089.2568 Iteration 1: log likelihood = -862.8039 Iteration 2: log likelihood = -847.99632 Iteration 3: log likelihood = -847.93063 Iteration 4: log likelihood = -847.93063 Ordered probit regression Number of obs = 1,020 Wald chi2(3) = 301.94 Log likelihood = -847.93063 Prob > chi2 = 0.0000 ( 1) [cut1]_cons + [cut2]_cons = 0 -------------------------------------------------------------------------------- res | Coef. Std. Err. z P>|z| [95% Conf. Interval] ---------------+---------------------------------------------------------------- difhomereal | .2952819 .0420416 7.02 0.000 .2128819 .3776819 ldiffifapoints | .940061 .0580157 16.20 0.000 .8263523 1.05377 ldifnumber | 2.186747 .8811128 2.48 0.013 .4597979 3.913697 ---------------+---------------------------------------------------------------- /cut1 | -.3896828 .0231681 -.4350915 -.3442742 /cut2 | .3896828 .0231681 .3442742 .4350915 -------------------------------------------------------------------------------- Note: 36 observations completely determined. Standard errors questionable. . estat ic, n(`games') Akaike's information criterion and Bayesian information criterion ----------------------------------------------------------------------------- Model | Obs ll(null) ll(model) df AIC BIC -------------+--------------------------------------------------------------- . | 1,020 . -847.9306 4 1703.861 1723.571 ----------------------------------------------------------------------------- Note: N=1020 used in calculating BIC. . predict lossfifa drawfifa winfifa (option pr assumed; predicted probabilities) . outreg2 using Table3a.xls, append keep(difhomereal ldiffifapoints ldifnumber) addstat(Log Likelihood, e(ll)) slow(750) dec(3) ctitle(F > IFA) Table3a.xls dir : seeout . . oprobit res difhomereal ldiffifapoints if selector==1, constraint(1) Iteration 0: log likelihood = -1089.2568 Iteration 1: log likelihood = -865.24114 Iteration 2: log likelihood = -851.17444 Iteration 3: log likelihood = -851.12078 Iteration 4: log likelihood = -851.12078 Ordered probit regression Number of obs = 1,020 Wald chi2(2) = 301.04 Log likelihood = -851.12078 Prob > chi2 = 0.0000 ( 1) [cut1]_cons + [cut2]_cons = 0 -------------------------------------------------------------------------------- res | Coef. Std. Err. z P>|z| [95% Conf. Interval] ---------------+---------------------------------------------------------------- difhomereal | .3050979 .0417921 7.30 0.000 .2231869 .3870088 ldiffifapoints | .9592866 .057427 16.70 0.000 .8467318 1.071841 ---------------+---------------------------------------------------------------- /cut1 | -.3881202 .0230852 -.4333665 -.3428739 /cut2 | .3881202 .0230852 .3428739 .4333665 -------------------------------------------------------------------------------- Note: 37 observations completely determined. Standard errors questionable. . estat ic, n(`games') Akaike's information criterion and Bayesian information criterion ----------------------------------------------------------------------------- Model | Obs ll(null) ll(model) df AIC BIC -------------+--------------------------------------------------------------- . | 1,020 . -851.1208 3 1708.242 1723.024 ----------------------------------------------------------------------------- Note: N=1020 used in calculating BIC. . outreg2 using Table3a.xls, append keep(difhomereal ldiffifapoints) addstat(Log Likelihood, e(ll)) slow(750) dec(3) ctitle(FIFA) Table3a.xls dir : seeout . . *ELO points . oprobit res difhomereal ldifelopoints ldifnumber if selector==1, constraint(1) Iteration 0: log likelihood = -1089.2568 Iteration 1: log likelihood = -832.17539 Iteration 2: log likelihood = -831.62743 Iteration 3: log likelihood = -831.62708 Iteration 4: log likelihood = -831.62708 Ordered probit regression Number of obs = 1,020 Wald chi2(3) = 359.69 Log likelihood = -831.62708 Prob > chi2 = 0.0000 ( 1) [cut1]_cons + [cut2]_cons = 0 ------------------------------------------------------------------------------- res | Coef. Std. Err. z P>|z| [95% Conf. Interval] --------------+---------------------------------------------------------------- difhomereal | .2832198 .0421165 6.72 0.000 .200673 .3657665 ldifelopoints | 4.498403 .2517499 17.87 0.000 4.004982 4.991824 ldifnumber | 1.337434 .8745837 1.53 0.126 -.3767188 3.051586 --------------+---------------------------------------------------------------- /cut1 | -.4003439 .0237771 -.4469462 -.3537417 /cut2 | .4003439 .0237771 .3537417 .4469462 ------------------------------------------------------------------------------- . estat ic, n(`games') Akaike's information criterion and Bayesian information criterion ----------------------------------------------------------------------------- Model | Obs ll(null) ll(model) df AIC BIC -------------+--------------------------------------------------------------- . | 1,020 . -831.6271 4 1671.254 1690.964 ----------------------------------------------------------------------------- Note: N=1020 used in calculating BIC. . predict losselo drawelo winelo (option pr assumed; predicted probabilities) . outreg2 using Table3a.xls, append keep(difhomereal ldifelopoints ldifnumber) addstat(Log Likelihood, e(ll)) slow(750) dec(3) ctitle(EL > O) Table3a.xls dir : seeout . . oprobit res difhomereal ldifelopoints if selector==1, constraint(1) Iteration 0: log likelihood = -1089.2568 Iteration 1: log likelihood = -833.18881 Iteration 2: log likelihood = -832.82383 Iteration 3: log likelihood = -832.82374 Iteration 4: log likelihood = -832.82374 Ordered probit regression Number of obs = 1,020 Wald chi2(2) = 360.59 Log likelihood = -832.82374 Prob > chi2 = 0.0000 ( 1) [cut1]_cons + [cut2]_cons = 0 ------------------------------------------------------------------------------- res | Coef. Std. Err. z P>|z| [95% Conf. Interval] --------------+---------------------------------------------------------------- difhomereal | .2892791 .0419127 6.90 0.000 .2071317 .3714264 ldifelopoints | 4.572021 .2473595 18.48 0.000 4.087205 5.056837 --------------+---------------------------------------------------------------- /cut1 | -.3997509 .023746 -.4462922 -.3532096 /cut2 | .3997509 .023746 .3532096 .4462922 ------------------------------------------------------------------------------- . estat ic, n(`games') Akaike's information criterion and Bayesian information criterion ----------------------------------------------------------------------------- Model | Obs ll(null) ll(model) df AIC BIC -------------+--------------------------------------------------------------- . | 1,020 . -832.8237 3 1671.647 1686.43 ----------------------------------------------------------------------------- Note: N=1020 used in calculating BIC. . outreg2 using Table3a.xls, append keep(difhomereal ldifelopoints) addstat(Log Likelihood, e(ll)) slow(750) dec(3) ctitle(ELO) Table3a.xls dir : seeout . . ***************************************GOAL MARGIN MODEL************************************* . *baseline . reg margin difhomereal ldifavvalue ldifnumber if selector==1 Source | SS df MS Number of obs = 1,020 -------------+---------------------------------- F(3, 1016) = 315.39 Model | 2443.92108 3 814.640362 Prob > F = 0.0000 Residual | 2624.27009 1,016 2.582943 R-squared = 0.4822 -------------+---------------------------------- Adj R-squared = 0.4807 Total | 5068.19118 1,019 4.97369105 Root MSE = 1.6072 ------------------------------------------------------------------------------ margin | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- difhomereal | .407706 .0523324 7.79 0.000 .3050141 .5103979 ldifavvalue | .5518693 .0200112 27.58 0.000 .5126014 .5911372 ldifnumber | 2.322923 .9178967 2.53 0.012 .521733 4.124113 _cons | .0875445 .0503259 1.74 0.082 -.0112101 .1862991 ------------------------------------------------------------------------------ . estat ic, n(`games') Akaike's information criterion and Bayesian information criterion ----------------------------------------------------------------------------- Model | Obs ll(null) ll(model) df AIC BIC -------------+--------------------------------------------------------------- . | 1,020 -2264.94 -1929.267 4 3866.535 3886.245 ----------------------------------------------------------------------------- Note: N=1020 used in calculating BIC. . outreg2 using Table3a.xls, append slow(750) dec(3) ctitle(Crowd) Table3a.xls dir : seeout . . reg margin difhomereal ldiffifapoints ldifnumber if selector==1 Source | SS df MS Number of obs = 1,020 -------------+---------------------------------- F(3, 1016) = 259.63 Model | 2199.34899 3 733.116329 Prob > F = 0.0000 Residual | 2868.84219 1,016 2.82366357 R-squared = 0.4340 -------------+---------------------------------- Adj R-squared = 0.4323 Total | 5068.19118 1,019 4.97369105 Root MSE = 1.6804 -------------------------------------------------------------------------------- margin | Coef. Std. Err. t P>|t| [95% Conf. Interval] ---------------+---------------------------------------------------------------- difhomereal | .4314511 .0547453 7.88 0.000 .3240244 .5388778 ldiffifapoints | .840606 .0340603 24.68 0.000 .7737694 .9074427 ldifnumber | 3.470062 .9522214 3.64 0.000 1.601516 5.338607 _cons | .0832044 .0526204 1.58 0.114 -.0200527 .1864615 -------------------------------------------------------------------------------- . estat ic, n(`games') Akaike's information criterion and Bayesian information criterion ----------------------------------------------------------------------------- Model | Obs ll(null) ll(model) df AIC BIC -------------+--------------------------------------------------------------- . | 1,020 -2264.94 -1974.711 4 3957.423 3977.133 ----------------------------------------------------------------------------- Note: N=1020 used in calculating BIC. . outreg2 using Table3a.xls, append slow(750) dec(3) ctitle(FIFA) Table3a.xls dir : seeout . . reg margin difhomereal ldiffifapoints if selector==1 Source | SS df MS Number of obs = 1,020 -------------+---------------------------------- F(2, 1017) = 378.24 Model | 2161.85069 2 1080.92535 Prob > F = 0.0000 Residual | 2906.34049 1,017 2.85775859 R-squared = 0.4266 -------------+---------------------------------- Adj R-squared = 0.4254 Total | 5068.19118 1,019 4.97369105 Root MSE = 1.6905 -------------------------------------------------------------------------------- margin | Coef. Std. Err. t P>|t| [95% Conf. Interval] ---------------+---------------------------------------------------------------- difhomereal | .451606 .054793 8.24 0.000 .3440857 .5591263 ldiffifapoints | .8723595 .0331251 26.34 0.000 .8073581 .9373609 _cons | .0846902 .0529355 1.60 0.110 -.0191852 .1885656 -------------------------------------------------------------------------------- . estat ic, n(`games') Akaike's information criterion and Bayesian information criterion ----------------------------------------------------------------------------- Model | Obs ll(null) ll(model) df AIC BIC -------------+--------------------------------------------------------------- . | 1,020 -2264.94 -1981.334 3 3968.669 3983.451 ----------------------------------------------------------------------------- Note: N=1020 used in calculating BIC. . outreg2 using Table3a.xls, append slow(750) dec(3) ctitle(FIFA) Table3a.xls dir : seeout . . reg margin difhomereal ldifelopoints ldifnumber if selector==1 Source | SS df MS Number of obs = 1,020 -------------+---------------------------------- F(3, 1016) = 339.92 Model | 2538.79224 3 846.264079 Prob > F = 0.0000 Residual | 2529.39894 1,016 2.48956589 R-squared = 0.5009 -------------+---------------------------------- Adj R-squared = 0.4995 Total | 5068.19118 1,019 4.97369105 Root MSE = 1.5778 ------------------------------------------------------------------------------- margin | Coef. Std. Err. t P>|t| [95% Conf. Interval] --------------+---------------------------------------------------------------- difhomereal | .4291231 .0513961 8.35 0.000 .3282685 .5299777 ldifelopoints | 6.338293 .2203794 28.76 0.000 5.905842 6.770744 ldifnumber | 1.528462 .907523 1.68 0.092 -.2523722 3.309296 _cons | .1156543 .0494115 2.34 0.019 .0186942 .2126145 ------------------------------------------------------------------------------- . estat ic, n(`games') Akaike's information criterion and Bayesian information criterion ----------------------------------------------------------------------------- Model | Obs ll(null) ll(model) df AIC BIC -------------+--------------------------------------------------------------- . | 1,020 -2264.94 -1910.489 4 3828.977 3848.687 ----------------------------------------------------------------------------- Note: N=1020 used in calculating BIC. . outreg2 using Table3a.xls, append slow(750) dec(3) ctitle(ELO) Table3a.xls dir : seeout . . reg margin difhomereal ldifelopoints if selector==1 Source | SS df MS Number of obs = 1,020 -------------+---------------------------------- F(2, 1017) = 507.55 Model | 2531.7304 2 1265.8652 Prob > F = 0.0000 Residual | 2536.46077 1,017 2.49406172 R-squared = 0.4995 -------------+---------------------------------- Adj R-squared = 0.4985 Total | 5068.19118 1,019 4.97369105 Root MSE = 1.5793 ------------------------------------------------------------------------------- margin | Coef. Std. Err. t P>|t| [95% Conf. Interval] --------------+---------------------------------------------------------------- difhomereal | .4377838 .0511843 8.55 0.000 .3373449 .5382228 ldifelopoints | 6.451396 .2100877 30.71 0.000 6.039141 6.863651 _cons | .1168465 .049451 2.36 0.018 .0198088 .2138841 ------------------------------------------------------------------------------- . estat ic, n(`games') Akaike's information criterion and Bayesian information criterion ----------------------------------------------------------------------------- Model | Obs ll(null) ll(model) df AIC BIC -------------+--------------------------------------------------------------- . | 1,020 -2264.94 -1911.911 3 3829.821 3844.604 ----------------------------------------------------------------------------- Note: N=1020 used in calculating BIC. . outreg2 using Table3a.xls, append slow(750) dec(3) ctitle(ELO) Table3a.xls dir : seeout . . *******************************************BIASES BY NUMBER OF FANS?************************************** . *look at correlation of bias measures . correl ldifavvalue ldifnumber ldifsupport playteamrelfans ldifnumbersupported ldifsupportvalue (obs=2,040) | ldifav~e ldifnu~r ldifsu~t pl~lfans ldifnu~d ldifsu~e -------------+------------------------------------------------------ ldifavvalue | 1.0000 ldifnumber | 0.2825 1.0000 ldifsupport | 0.7404 0.1756 1.0000 playte~lfans | 0.2626 0.0072 0.4154 1.0000 ldifnumber~d | 0.5962 0.1386 0.8481 0.4859 1.0000 ldifsuppor~e | 0.8428 0.2256 0.9509 0.3689 0.7765 1.0000 . . *game outcome model . oprobit res difhomereal ldifavvalue ldifnumber ldifsupport if selector==1, constraint(1) Iteration 0: log likelihood = -1089.2568 Iteration 1: log likelihood = -830.69899 Iteration 2: log likelihood = -820.22495 Iteration 3: log likelihood = -820.17162 Iteration 4: log likelihood = -820.17162 Ordered probit regression Number of obs = 1,020 Wald chi2(4) = 377.88 Log likelihood = -820.17162 Prob > chi2 = 0.0000 ( 1) [cut1]_cons + [cut2]_cons = 0 ------------------------------------------------------------------------------ res | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- difhomereal | .2806672 .0422424 6.64 0.000 .1978735 .3634609 ldifavvalue | .4474967 .0356293 12.56 0.000 .3776646 .5173288 ldifnumber | 1.778153 .8800568 2.02 0.043 .0532738 3.503033 ldifsupport | .0025368 .0226541 0.11 0.911 -.0418644 .0469381 -------------+---------------------------------------------------------------- /cut1 | -.4088815 .0242954 -.4564996 -.3612634 /cut2 | .4088815 .0242954 .3612634 .4564996 ------------------------------------------------------------------------------ Note: 7 observations completely determined. Standard errors questionable. . estat ic, n(`games') Akaike's information criterion and Bayesian information criterion ----------------------------------------------------------------------------- Model | Obs ll(null) ll(model) df AIC BIC -------------+--------------------------------------------------------------- . | 1,020 . -820.1716 5 1650.343 1674.981 ----------------------------------------------------------------------------- Note: N=1020 used in calculating BIC. . outreg2 using Table8.xls, append keep(difhomereal ldifavvalue ldifnumber ldifsupport) addstat(Log Likelihood, e(ll)) slow(750) dec(3) > ctitle(game result) Table8.xls dir : seeout . . oprobit res difhomereal ldifavvalue ldifnumber difplayteamrelfans if selector==1, constraint(1) Iteration 0: log likelihood = -1089.2568 Iteration 1: log likelihood = -822.08968 Iteration 2: log likelihood = -816.4735 Iteration 3: log likelihood = -816.45068 Iteration 4: log likelihood = -816.45068 Ordered probit regression Number of obs = 1,020 Wald chi2(4) = 385.03 Log likelihood = -816.45068 Prob > chi2 = 0.0000 ( 1) [cut1]_cons + [cut2]_cons = 0 ------------------------------------------------------------------------------------ res | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------------+---------------------------------------------------------------- difhomereal | .2798297 .0422716 6.62 0.000 .1969788 .3626806 ldifavvalue | .42194 .0263297 16.03 0.000 .3703347 .4735452 ldifnumber | 2.075861 .8976223 2.31 0.021 .3165536 3.835168 difplayteamrelfans | 10.01182 3.764545 2.66 0.008 2.633445 17.39019 -------------------+---------------------------------------------------------------- /cut1 | -.4113958 .0244403 -.459298 -.3634936 /cut2 | .4113958 .0244403 .3634936 .459298 ------------------------------------------------------------------------------------ Note: 7 observations completely determined. Standard errors questionable. . estat ic, n(`games') Akaike's information criterion and Bayesian information criterion ----------------------------------------------------------------------------- Model | Obs ll(null) ll(model) df AIC BIC -------------+--------------------------------------------------------------- . | 1,020 . -816.4507 5 1642.901 1667.539 ----------------------------------------------------------------------------- Note: N=1020 used in calculating BIC. . outreg2 using Table8.xls, append keep(difhomereal ldifavvalue ldifnumber difplayteamrelfans) addstat(Log Likelihood, e(ll)) slow(750) > dec(3) ctitle(game result) Table8.xls dir : seeout . . oprobit res difhomereal ldifavvalue ldifnumber ldifnumbersupported if selector==1, constraint(1) Iteration 0: log likelihood = -1089.2568 Iteration 1: log likelihood = -829.07225 Iteration 2: log likelihood = -819.83081 Iteration 3: log likelihood = -819.78192 Iteration 4: log likelihood = -819.78192 Ordered probit regression Number of obs = 1,020 Wald chi2(4) = 380.57 Log likelihood = -819.78192 Prob > chi2 = 0.0000 ( 1) [cut1]_cons + [cut2]_cons = 0 ------------------------------------------------------------------------------------- res | Coef. Std. Err. z P>|z| [95% Conf. Interval] --------------------+---------------------------------------------------------------- difhomereal | .2808459 .0422334 6.65 0.000 .1980698 .3636219 ldifavvalue | .4333795 .030841 14.05 0.000 .3729323 .4938268 ldifnumber | 1.788638 .8793561 2.03 0.042 .0651313 3.512144 ldifnumbersupported | .0438516 .0492714 0.89 0.373 -.0527185 .1404216 --------------------+---------------------------------------------------------------- /cut1 | -.4092994 .0243233 -.4569721 -.3616267 /cut2 | .4092994 .0243233 .3616267 .4569721 ------------------------------------------------------------------------------------- Note: 7 observations completely determined. Standard errors questionable. . estat ic, n(`games') Akaike's information criterion and Bayesian information criterion ----------------------------------------------------------------------------- Model | Obs ll(null) ll(model) df AIC BIC -------------+--------------------------------------------------------------- . | 1,020 . -819.7819 5 1649.564 1674.202 ----------------------------------------------------------------------------- Note: N=1020 used in calculating BIC. . outreg2 using Table8.xls, append keep(difhomereal ldifavvalue ldifnumber ldifnumbersupported) addstat(Log Likelihood, e(ll)) slow(750) > dec(3) ctitle(game result) Table8.xls dir : seeout . predict losssup drawsup winsup (option pr assumed; predicted probabilities) . . oprobit res difhomereal ldifavvalue ldifnumber ldifsupportvalue if selector==1, constraint(1) Iteration 0: log likelihood = -1089.2568 Iteration 1: log likelihood = -833.03365 Iteration 2: log likelihood = -819.99269 Iteration 3: log likelihood = -819.92486 Iteration 4: log likelihood = -819.92485 Ordered probit regression Number of obs = 1,020 Wald chi2(4) = 376.99 Log likelihood = -819.92485 Prob > chi2 = 0.0000 ( 1) [cut1]_cons + [cut2]_cons = 0 ---------------------------------------------------------------------------------- res | Coef. Std. Err. z P>|z| [95% Conf. Interval] -----------------+---------------------------------------------------------------- difhomereal | .2805859 .0422512 6.64 0.000 .197775 .3633967 ldifavvalue | .4788409 .0469354 10.20 0.000 .3868492 .5708325 ldifnumber | 1.734926 .881249 1.97 0.049 .0077098 3.462142 ldifsupportvalue | -.0126959 .0178458 -0.71 0.477 -.047673 .0222812 -----------------+---------------------------------------------------------------- /cut1 | -.4091122 .0243111 -.4567611 -.3614632 /cut2 | .4091122 .0243111 .3614632 .4567611 ---------------------------------------------------------------------------------- Note: 7 observations completely determined. Standard errors questionable. . estat ic, n(`games') Akaike's information criterion and Bayesian information criterion ----------------------------------------------------------------------------- Model | Obs ll(null) ll(model) df AIC BIC -------------+--------------------------------------------------------------- . | 1,020 . -819.9249 5 1649.85 1674.487 ----------------------------------------------------------------------------- Note: N=1020 used in calculating BIC. . outreg2 using Table8.xls, append keep(difhomereal ldifavvalue ldifnumber ldifsupportvalue) addstat(Log Likelihood, e(ll)) slow(750) de > c(3) ctitle(game result) Table8.xls dir : seeout . . ***Goal margin model . reg margin difhomereal ldifavvalue ldifnumber if selector==1 Source | SS df MS Number of obs = 1,020 -------------+---------------------------------- F(3, 1016) = 315.39 Model | 2443.92108 3 814.640362 Prob > F = 0.0000 Residual | 2624.27009 1,016 2.582943 R-squared = 0.4822 -------------+---------------------------------- Adj R-squared = 0.4807 Total | 5068.19118 1,019 4.97369105 Root MSE = 1.6072 ------------------------------------------------------------------------------ margin | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- difhomereal | .407706 .0523324 7.79 0.000 .3050141 .5103979 ldifavvalue | .5518693 .0200112 27.58 0.000 .5126014 .5911372 ldifnumber | 2.322923 .9178967 2.53 0.012 .521733 4.124113 _cons | .0875445 .0503259 1.74 0.082 -.0112101 .1862991 ------------------------------------------------------------------------------ . estat ic, n(`games') Akaike's information criterion and Bayesian information criterion ----------------------------------------------------------------------------- Model | Obs ll(null) ll(model) df AIC BIC -------------+--------------------------------------------------------------- . | 1,020 -2264.94 -1929.267 4 3866.535 3886.245 ----------------------------------------------------------------------------- Note: N=1020 used in calculating BIC. . outreg2 using Table8.xls, append keep(difhomereal ldifavvalue ldifnumber) slow(750) dec(3) ctitle(goal margin) Table8.xls dir : seeout . . reg margin difhomereal ldifavvalue ldifnumber ldifsupport if selector==1 Source | SS df MS Number of obs = 1,020 -------------+---------------------------------- F(4, 1015) = 239.10 Model | 2458.78465 4 614.696163 Prob > F = 0.0000 Residual | 2609.40653 1,015 2.57084387 R-squared = 0.4851 -------------+---------------------------------- Adj R-squared = 0.4831 Total | 5068.19118 1,019 4.97369105 Root MSE = 1.6034 ------------------------------------------------------------------------------ margin | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- difhomereal | .4062267 .0522133 7.78 0.000 .3037684 .5086851 ldifavvalue | .5003506 .0292856 17.09 0.000 .4428834 .5578178 ldifnumber | 2.440385 .9170464 2.66 0.008 .6408613 4.239909 ldifsupport | .0605707 .0251906 2.40 0.016 .011139 .1100024 _cons | .0833802 .0502378 1.66 0.097 -.0152016 .1819619 ------------------------------------------------------------------------------ . estat ic, n(`games') Akaike's information criterion and Bayesian information criterion ----------------------------------------------------------------------------- Model | Obs ll(null) ll(model) df AIC BIC -------------+--------------------------------------------------------------- . | 1,020 -2264.94 -1926.371 5 3862.741 3887.379 ----------------------------------------------------------------------------- Note: N=1020 used in calculating BIC. . outreg2 using Table8.xls, append keep(difhomereal ldifavvalue ldifnumber ldifsupport) slow(750) dec(3) ctitle(goal margin) Table8.xls dir : seeout . . reg margin difhomereal ldifavvalue ldifnumber difplayteamrelfans if selector==1 Source | SS df MS Number of obs = 1,020 -------------+---------------------------------- F(4, 1015) = 244.27 Model | 2485.84734 4 621.461836 Prob > F = 0.0000 Residual | 2582.34383 1,015 2.54418111 R-squared = 0.4905 -------------+---------------------------------- Adj R-squared = 0.4885 Total | 5068.19118 1,019 4.97369105 Root MSE = 1.595 ------------------------------------------------------------------------------------ margin | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------------+---------------------------------------------------------------- difhomereal | .4065478 .051939 7.83 0.000 .3046277 .508468 ldifavvalue | .5180946 .0215327 24.06 0.000 .4758408 .5603484 ldifnumber | 2.723958 .9163241 2.97 0.003 .9258514 4.522064 difplayteamrelfans | 16.44373 4.050712 4.06 0.000 8.495005 24.39246 _cons | .0724002 .050086 1.45 0.149 -.0258838 .1706841 ------------------------------------------------------------------------------------ . estat ic, n(`games') Akaike's information criterion and Bayesian information criterion ----------------------------------------------------------------------------- Model | Obs ll(null) ll(model) df AIC BIC -------------+--------------------------------------------------------------- . | 1,020 -2264.94 -1921.054 5 3852.107 3876.745 ----------------------------------------------------------------------------- Note: N=1020 used in calculating BIC. . outreg2 using Table8.xls, append keep(difhomereal ldifavvalue ldifnumber difplayteamrelfans) slow(750) dec(3) ctitle(goal margin) Table8.xls dir : seeout . . reg margin difhomereal ldifavvalue ldifnumber ldifnumbersupported if selector==1 Source | SS df MS Number of obs = 1,020 -------------+---------------------------------- F(4, 1015) = 239.31 Model | 2459.86078 4 614.965196 Prob > F = 0.0000 Residual | 2608.33039 1,015 2.56978364 R-squared = 0.4854 -------------+---------------------------------- Adj R-squared = 0.4833 Total | 5068.19118 1,019 4.97369105 Root MSE = 1.6031 ------------------------------------------------------------------------------------- margin | Coef. Std. Err. t P>|t| [95% Conf. Interval] --------------------+---------------------------------------------------------------- difhomereal | .408251 .0521994 7.82 0.000 .3058199 .510682 ldifavvalue | .5158981 .0246376 20.94 0.000 .4675516 .5642446 ldifnumber | 2.410088 .9162242 2.63 0.009 .6121778 4.207998 ldifnumbersupported | .1389994 .0558112 2.49 0.013 .0294809 .2485179 _cons | .0869142 .0501982 1.73 0.084 -.0115899 .1854183 ------------------------------------------------------------------------------------- . estat ic, n(`games') Akaike's information criterion and Bayesian information criterion ----------------------------------------------------------------------------- Model | Obs ll(null) ll(model) df AIC BIC -------------+--------------------------------------------------------------- . | 1,020 -2264.94 -1926.16 5 3862.32 3886.958 ----------------------------------------------------------------------------- Note: N=1020 used in calculating BIC. . outreg2 using Table8.xls, append keep(difhomereal ldifavvalue ldifnumber ldifnumbersupported) slow(750) dec(3) ctitle(goal margin) Table8.xls dir : seeout . . reg margin difhomereal ldifavvalue ldifnumber ldifsupportvalue if selector==1 Source | SS df MS Number of obs = 1,020 -------------+---------------------------------- F(4, 1015) = 240.15 Model | 2464.31117 4 616.077792 Prob > F = 0.0000 Residual | 2603.88001 1,015 2.56539902 R-squared = 0.4862 -------------+---------------------------------- Adj R-squared = 0.4842 Total | 5068.19118 1,019 4.97369105 Root MSE = 1.6017 ---------------------------------------------------------------------------------- margin | Coef. Std. Err. t P>|t| [95% Conf. Interval] -----------------+---------------------------------------------------------------- difhomereal | .4080856 .0521545 7.82 0.000 .3057425 .5104286 ldifavvalue | .4669532 .0361241 12.93 0.000 .3960666 .5378397 ldifnumber | 2.385598 .9150442 2.61 0.009 .5900033 4.181193 ldifsupportvalue | .0497615 .0176507 2.82 0.005 .0151255 .0843975 _cons | .0805327 .0502163 1.60 0.109 -.018007 .1790724 ---------------------------------------------------------------------------------- . estat ic, n(`games') Akaike's information criterion and Bayesian information criterion ----------------------------------------------------------------------------- Model | Obs ll(null) ll(model) df AIC BIC -------------+--------------------------------------------------------------- . | 1,020 -2264.94 -1925.289 5 3860.579 3885.216 ----------------------------------------------------------------------------- Note: N=1020 used in calculating BIC. . outreg2 using Table8.xls, append keep(difhomereal ldifavvalue ldifnumber ldifsupportvalue) slow(750) dec(3) ctitle(goal margin) Table8.xls dir : seeout . . ********************************************COMPARE FIT OF MODELS AT THE MATCH LEVEL******************************** . keep if home==1 (1,020 observations deleted) . . egen lossnull=mean(loss) . egen drawnull=mean(draw) . egen winnull=mean(win) . . gen fitnull=winnull*win+drawnull*draw+lossnull*loss . gen fitodds=winodds*win+drawodds*draw+lossodds*loss . gen fitfifa=winfifa*win+drawfifa*draw+lossfifa*loss . gen fitelo=winelo*win+drawelo*draw+losselo*loss . gen fithat=winhat*win+drawhat*draw+losshat*loss . gen fitsup=winsup*win+drawsup*draw+losssup*loss . gen fitlogit=winlogit*win+drawlogit*draw+losslogit*loss . gen fitmlogit=winmlogit*win+drawmlogit*draw+lossmlogit*loss . . ***Geometric mean of fits . . ameans fit* Variable | Type Obs Mean [95% Conf. Interval] -------------+--------------------------------------------------------------- fitnull | Arithmetic 1,020 .3633583 .3572465 .3694701 | Geometric 1,020 .3486267 .3423601 .3550079 | Harmonic 1,020 .3333333 .3271263 .3397805 -------------+--------------------------------------------------------------- fitodds | Arithmetic 1,020 .5123937 .4973761 .5274113 | Geometric 1,020 .4461901 .4310336 .4618796 | Harmonic 1,020 .3725153 .3548111 .3920791 -------------+--------------------------------------------------------------- fitfifa | Arithmetic 1,020 .5098915 .4937851 .525998 | Geometric 1,020 .4354808 .4192347 .4523565 | Harmonic 1,020 .2707297 .1835089 .5159653 -------------+--------------------------------------------------------------- fitelo | Arithmetic 1,020 .5215015 .5050638 .5379393 | Geometric 1,020 .4424974 .425466 .4602106 | Harmonic 1,020 .2330507 .1372776 .7708237 -------------+--------------------------------------------------------------- fithat | Arithmetic 1,020 .5283572 .511792 .5449224 | Geometric 1,020 .4474923 .4303839 .4652807 | Harmonic 1,020 .3368042 .2989651 .3856095 -------------+--------------------------------------------------------------- fitsup | Arithmetic 1,020 .5286754 .5121112 .5452396 | Geometric 1,020 .447666 .4305353 .4654784 | Harmonic 1,020 .3376252 .3011431 .3841649 -------------+--------------------------------------------------------------- fitlogit | Arithmetic 1,020 .5277458 .5111891 .5443025 | Geometric 1,020 .4468343 .4298633 .4644754 | Harmonic 1,020 .3499906 .3270795 .3763532 -------------+--------------------------------------------------------------- fitmlogit | Arithmetic 1,020 .5235845 .5075873 .5395817 | Geometric 1,020 .4450916 .4280776 .4627818 | Harmonic 1,020 .329831 .2904299 .3816008 -------------+--------------------------------------------------------------- . . ***Implied Log likelihood . . gen lnull=log(fitnull) . gen lodds=log(fitodds) . gen lfifa=log(fitfifa) . gen lelo=log(fitelo) . gen lhat=log(fithat) . gen lsup=log(fitsup) . gen llogit=log(fitlogit) . gen lmlogit=log(fitmlogit) . . egen llnull=total(lnull) . egen llodds=total(lodds) . egen llfifa=total(lfifa) . egen llelo=total(lelo) . egen llhat=total(lhat) . egen llsup=total(lsup) . egen lllogit=total(llogit) . egen llmlogit=total(lmlogit) . . sum ll* Variable | Obs Mean Std. Dev. Min Max -------------+--------------------------------------------------------- llogit | 1,020 -.8055674 .6302016 -4.28362 -.0000105 llnull | 1,020 -1074.829 0 -1074.829 -1074.829 llodds | 1,020 -823.1504 0 -823.1504 -823.1504 llfifa | 1,020 -847.9306 0 -847.9306 -847.9306 llelo | 1,020 -831.6271 0 -831.6271 -831.6271 -------------+--------------------------------------------------------- llhat | 1,020 -820.1779 0 -820.1779 -820.1779 llsup | 1,020 -819.7819 0 -819.7819 -819.7819 lllogit | 1,020 -821.6787 0 -821.6787 -821.6787 llmlogit | 1,020 -825.6646 0 -825.6646 -825.6646 . . ***Corealtion between fitted values . corr winhat winsup winodds winfifa winelo (obs=1,020) | winhat winsup winodds winfifa winelo -------------+--------------------------------------------- winhat | 1.0000 winsup | 0.9993 1.0000 winodds | 0.9568 0.9574 1.0000 winfifa | 0.9214 0.9212 0.9402 1.0000 winelo | 0.9447 0.9448 0.9661 0.9657 1.0000 . corr drawhat drawsup drawodds drawfifa drawelo (obs=1,020) | drawhat drawsup drawodds drawfifa drawelo -------------+--------------------------------------------- drawhat | 1.0000 drawsup | 0.9987 1.0000 drawodds | 0.9019 0.9032 1.0000 drawfifa | 0.8535 0.8524 0.8702 1.0000 drawelo | 0.8944 0.8930 0.9260 0.9207 1.0000 . . log close name: log: C:\Users\56570tpe\Dropbox\work\transfer values\stata\replication files\allgamesindividuals.log log type: text closed on: 7 Aug 2017, 15:53:05 ----------------------------------------------------------------------------------------------------------------------------------------