Good evening everyone,
since emmeans is not yet compatible with the “categorical” family, I am trying to calculate the marginal effects with the solution described in link:
setting “conditions = data.frame(speaker = NA)”.
Despite I set conditions, the estimates don’t change and I’m afraid to calculate “their first level assigned” in both cases. All my predictors are factorial.
profile~TIME+coho+sex+field+(1|ID)+(1|Academie) is the formula of the model
version brms: 2.16.3
I found the error, I didn’t use the SUM CODING for dummies variables.
Then I applied it as explained here: https://phillipalday.com/stats/coding.html.
I generated new variables to use as dummies:
dati4=dati3
dati4$sex1[dati4$sex==“F”]=-1
dati4$sex1[dati4$sex==“M”]=1
dati4$TIME1[dati4$TIME==“1”]=-1
dati4$TIME1[dati4$TIME==“2”]=1
dati4$coho1[dati4$coho==“2016”]=-1
dati4$coho1[dati4$coho==“2020”]=1
dati4$field1[dati4$field==“A”]=-1
dati4$field2[dati4$field==“A”]=-1
dati4$field1[dati4$field==“B”]=1
dati4$field2[dati4$field==“B”]=0
dati4$field1[dati4$field==“C”]=0
dati4$field2[dati4$field==“C”]=1
then for every predictor of interest I made a specific regression model where the first variable is the one with the levels I want to analyse, and the others are the categorical variables in dummy version:
modelSex=brm(profile~sex+TIME1+coho1+field1+field2+(1|ID)+(1|Academie), data=dati4, family=“categorical”…
modelCoho=brm(profile~coho+sex1+TIME1+field1+field2+(1|ID)+(1|Academie), data=dati4, family=“categorical”…
modelField=brm(profile~field+sex1+TIME1+coho1+(1|ID)+(1|Academie), data=dati4, family=“categorical”…
modelTIME=brm(profile~TIME+sex1+coho1+field1+field2+(1|ID)+(1|Academie), data=dati4, family=“categorical”…
and obtained the marginals effects from from each of them:
conditional_effects(modelTIME, categorical=T)
conditional_effects(modelCoho, categorical=T)
conditional_effects(modelSex, categorical=T)
conditional_effects(modelField, categorical=T)
do you think this solution is reliable?
