Skip to content
Prev 278575 / 398502 Next

computationally singular error with mice()

Hi Fei,

I wouldn't worry to much about categorical variables for mice. Mice
would use logisitic regression for binary and polytomous logistic
regression for categorical variables with >2 levels. However, you
should not include factors with a lot of levels, saying>30, in
imputation models because it would require a lot of dummy variables.

Another thing is that not excluding variables you would use in
substantive analysis. Otherwise, estimation would be biased.

Weidong
On Sat, Nov 26, 2011 at 12:07 PM, Fei <fayechen0807 at hotmail.com> wrote: