0

I would like to include fixed effects in my Tobit regression in r. This regression has a flexible upper limit and is right-censored. This is the r code for the Tobit regression: 

R1 <- tobit(LNcAttendance ~ LNcCapacityH + LNcPrAttendanceH + LNcPrAttendanceA +
              TeamQuality_100 + CumSurprise_100 + HProb + HProb2 + ChampionshipS + 
              CLS + ELS + RelegationS + PMDRank + Temperature + Precipitation +
              Derby + PromotedH + Weekday + JulyAugust + September + October + 
              November + December + February + March + April + MayJune,
            left=-Inf, right=THESISDATA$UpperLimit, data=THESISDATA))

summary(R1)

I want to include team and calendar fixed effects in one regression and team-year fixed effects in a second regression. In my data frame, one observation represents a football match for which all variables in the Tobit regression above are collected plus calender year, hometeam, and awayteam. This is what I have tried so far: 

R141 <- tobit(LNcAttendance ~ LNcCapacityH + LNcPrAttendanceH + LNcPrAttendanceA + 
                TeamQuality_100 + CumSurprise_100 + HProb + HProb2 + ChampionshipS +
                CLS + ELS + RelegationS + PMDRank + Temperature + Precipitation +
                Derby + PromotedH + Weekday + JulyAugust + September + October + 
                November + December + February + March + April + MayJune +  
                factor(HomeTeam)-1 + factor(Year)-1,left=-Inf,
              right=THESISDATA$UpperLimit, data=THESISDATA)

and

R142 <- plm(LNcAttendance ~ LNcCapacityH + LNcPrAttendanceH + LNcPrAttendanceA + 
              TeamQuality_100 + CumSurprise_100 + HProb + HProb2 + ChampionshipS + 
              CLS + ELS + RelegationS + PMDRank + Temperature + Precipitation +
              Derby + PromotedH + Weekday + JulyAugust + September + October + 
              November + December + February + March + April + MayJune,
            left=-Inf, right=THESISDATA$UpperLimit, data=THESISDATA, model="whitin")
jay.sf
  • 60,139
  • 8
  • 53
  • 110
Kim van Dijk
  • 1
  • 1
  • Looks a bit like overfitting; regardless we don't know where the `tobit` function stems from, and don't have `THESISDATA` data. Overall we need a minimal self-contained example to answer questions, as explained here: https://stackoverflow.com/questions/5963269/how-to-make-a-great-r-reproducible-example/5963610#5963610 – jay.sf May 16 '20 at 17:03
  • Thanks for the feedback. The model I used is a copy of an existing paper. The first regression (R1) tries to determine the variables that affect stadium attendance in football. It is assumed that the stadium is full at 95% of the capacity, and this is why a Tobit model is used. For sensitivity analysis, I want to see if team and calendar year fixed effects and team-year fixed effects change the results. This is what THESISDATA looks like: – Kim van Dijk May 17 '20 at 07:44
  • > Date Year Season HomeTeam AwayTeam ... > > 2013-08-02 2013 2013/~ Ajax Roda ... > 2013-08-03 2013 2013/~ Den Haag PSV Ein~ ... > This are only 2 observations, but in total THESISDATA contains 1,836 observations. The variables in the dataset are: Date, Year, Season, HomeTeam, AwayTeam, LNcCapacityH, LNcAttendance, LNcPrAttendanceH, LNcPrAttendanceA, UpperLimit, TeamQuality_100, CumSurprise_100, ... (all the variables in the regressions above) – Kim van Dijk May 17 '20 at 07:55

0 Answers0