I fit my data with linear mix model using y and log-transformed y like below:

fit1 = lmer(y~Visit+Treatment+(1|Subject.ID))
fit2 = lmer(log(y)~Visit+Treatment+(1|Subject.ID))

Treatment effect in fit1 is not significant, but it is significant in fit2 (with log(y) ).

I'm confused about how should I interpret the model result.

Can anyone explain what is the correct way to interpret mixed model result when dependent variable is log-transformed vs not log-transformed.

Thanks a lot in advance!

  • 2
    $\begingroup$ This has come up before. One example here and another here. $\endgroup$
    – dimitriy
    Commented Jan 15, 2019 at 23:02
  • $\begingroup$ Here is another possible duplicate. Also, it would be a good idea to check that the residuals are normally distributed. $\endgroup$ Commented Jan 16, 2019 at 11:22


Browse other questions tagged or ask your own question.