I did multiple imputation with mice in R. My outcome model includes an interaction term between two categorical variables (predictor: gender 0:1; moderator: poverty 1:2:3). For this, I tried to split a dataset into three datasets (by poverty group) and then impute each dataset separately. Then, I combined the imputed datasets in order to run final outcome analyses.
However, the results from the outcome analyses did not show the interaction coefficients correctly.. Here is my code:
#convert to factor
data$pov <- as.factor(data$pov)
#split dataset by poverty group
data_pov1=subset(data, pov==1)
data_pov2=subset(data, pov==2)
data_pov3=subset(data, pov==3)
#impute each dataset
imp.pov1 <- mice(data=data_pov1, m=18, seed=12345, print=FALSE)
imp.pov2 <- mice(data=data_pov2, m=18, seed=12345, print=FALSE)
imp.pov3 <- mice(data=data_pov3, m=18, seed=12345, print=FALSE)
#combine each imputed dataset
imp.pov.t <- rbind(imp.pov1, imp.pov2, imp.pov3)
#final analyses
mi_pov <- with(imp.pov.t, lm(saf~ gender + pov + gender*pov))
pool.fit <-pool(mi_pov)
summary(pool.fit)
From the final analyses, I got the result like this: gender XXX pov2 XXX gender:pov2 XXX
I do not know why the results include the coefficients of pov2 and gender:pov2. I would like to get the coefficients of pov2, pov3, gender:pov2, and gender:pov3 (ref=pov1). I am very new to R.. I would be grateful if anyone can offer me some help. I appreciate your help in advance.
