I have a problem with calculating the model fit for different models calculated with the fmlogit and margins command.
I'm using the fmlogit package in stata 13 for windows calculating a multinomial fractional logit model with three different dependent variables (all values between 0 and 1). The case numbe ris constant by 380. The basic model includes independent variables from the individual and contextual level. I then calculate two other fmlogit-models. The first includes only variables from the individual level and the second model includes only variables from the contextual level.
I then calcualte for each model (3) and dependent variables (3) the margins for a better interpretation of my variables.
My problem is the following: I have to compare the different models (including independet variables from (1) the contextual level, (2) the individual level and (3) both levels) and present something like an model fit. But all the common postestimation comands are not working.
Is there something like AIC/BIC, lrtest, lntest I could possibly run to proof the validity and model fit?
I tried the fmlogit postestimation command but it doesnt work.
fmlogit fraclogrep fraclogpart fraclogdirekt, eta(gender alter2 bildungmax zuwanderung2 wohnungsbau2 ///
parteilos cdu E2 kohabitation1 mehrheitspartei formell ///
informell popkm veraenderung14_94 vereindichte haushalt2 arbeitsquote2017)
margins, dydx(*) post predict(outcome(fraclogrep)) // marginal effects for representation on the contextual AND individual level
fmlogit fraclogrep fraclogpart fraclogdirekt, eta(E2 popkm veraenderung14_94 vereindichte haushalt2 arbeitsquote2017)
margins, dydx(*) post predict(outcome(fraclogrep)) // marginal effects for representation on the contextual level (ONLY)
fmlogit fraclogrep fraclogpart fraclogdirekt, eta(gender alter2 bildungmax zuwanderung2 wohnungsbau2 ///
parteilos cdu kohabitation1 mehrheitspartei formell ///
informell)
margins, dydx(*) post predict(outcome(fraclogrep)) // marginal effects for representation on the individual level (ONLY)
I'm using the fmlogit package in stata 13 for windows calculating a multinomial fractional logit model with three different dependent variables (all values between 0 and 1). The case numbe ris constant by 380. The basic model includes independent variables from the individual and contextual level. I then calculate two other fmlogit-models. The first includes only variables from the individual level and the second model includes only variables from the contextual level.
I then calcualte for each model (3) and dependent variables (3) the margins for a better interpretation of my variables.
My problem is the following: I have to compare the different models (including independet variables from (1) the contextual level, (2) the individual level and (3) both levels) and present something like an model fit. But all the common postestimation comands are not working.
Is there something like AIC/BIC, lrtest, lntest I could possibly run to proof the validity and model fit?
I tried the fmlogit postestimation command but it doesnt work.
fmlogit fraclogrep fraclogpart fraclogdirekt, eta(gender alter2 bildungmax zuwanderung2 wohnungsbau2 ///
parteilos cdu E2 kohabitation1 mehrheitspartei formell ///
informell popkm veraenderung14_94 vereindichte haushalt2 arbeitsquote2017)
margins, dydx(*) post predict(outcome(fraclogrep)) // marginal effects for representation on the contextual AND individual level
fmlogit fraclogrep fraclogpart fraclogdirekt, eta(E2 popkm veraenderung14_94 vereindichte haushalt2 arbeitsquote2017)
margins, dydx(*) post predict(outcome(fraclogrep)) // marginal effects for representation on the contextual level (ONLY)
fmlogit fraclogrep fraclogpart fraclogdirekt, eta(gender alter2 bildungmax zuwanderung2 wohnungsbau2 ///
parteilos cdu kohabitation1 mehrheitspartei formell ///
informell)
margins, dydx(*) post predict(outcome(fraclogrep)) // marginal effects for representation on the individual level (ONLY)
