错误边界（单数）拟合：参见 lme4 Model 中的 isSingular

Question

I have the following structure:

'data.frame':   1041 obs. of  4 variables:
 $ RST_AST: num  0.45 0.576 0.45 0.45 0.675 ...
 $ subj   : chr  "4" "13" "24" "25" ...
 $ domain : chr  "perceptual" "perceptual" "perceptual" "perceptual" ...
 $ ce     : chr  "compromise" "compromise" "compromise" "compromise" ...

Now I want to run a Multilevel Model using lme4 package like these:

library(lme4) #load package

ce_model <- lmer(RST_AST ~ domain + ce+ (1|subj), data = df_RST_AST)

Then I get the error message:

boundary (singular) fit: see ?isSingular

I have no clue what's wrong?

Can someone help me?

Answer 1

It is likely that one of your variables is not being evaluated as contributing to the variance. Most likely, the "culprit" is the random variable (subj in your case). If you look at the output of your model (just print ce_model and hit enter, or summary(ce_model)), there should be a variable that has zero or very nearly zero variance and standard deviation. There are several ways to deal with this (see any of the links above). I would recommend analyzing the faulty variable, what kind of distribution does it have, can it be transformed, ...etc. Perhaps you need to do one or more generalized models with the data. It might also be helpful to look at your model or models; however, without further info it is hard to tell.