proc mi warning An effect for variable X1 is a linear combination of other effects.

emilyyi — Tue, 08 Mar 2016 20:07:52 GMT

Hi everyone,

I am using MICE to do the multiple imputation. The independent variables include both numeric (X1) and categorical variables (X2, X3). The dependent variable is a 3 level variable. The code as follow:

proc mi data=testa seed=123 nimpute = 5 out=outfcs;
class Y X2 X3;
fcs nbiter=10 discrim(Y X2 X3 /details classeffects = include) ;
var X1 X2 X3 Y;
run;

It gives warning: An effect for variable X1 is a linear combination of other effects. The coefficient of the effect will be set
to zero in the imputation.

However, if I remove X1 from the MI process, it gives the error: Each observation has analysis variables either all missing or all observed in the data set.

I searched a lot, but I did not see many topics on this warning in the proc mi. Anyone help me find a solution to it?

Thanks so much!

Re: proc mi warning An effect for variable X1 is a linear combination of other effects.

Ksharp — Wed, 09 Mar 2016 01:07:10 GMT

You do not need to remove X1 , SAS will set its parameter to zero defaultly . It is like so called multi-colinear in PROC REG .

topic Re: proc mi warning An effect for variable X1 is a linear combination of other effects. in Statistical Procedures

proc mi warning An effect for variable X1 is a linear combination of other effects.

Re: proc mi warning An effect for variable X1 is a linear combination of other effects.