Dear all,
I want to study the effects of several factors on a quantitative variable with aggregated data (several observations per unit). Therefore I have performed a linear mixed-effects model. The residuals are not normally distributed (significant normality test) but the QQ-plot is quite satisfactory, except the fact it shows a slightly right-skewed distribution. I precise that before performing the linear mixed-effects model, I have transformed the dependant variable with the square root function and the ln transformation does not work. Because of the non-normality the tests for fixed effects are not reliable. Is it possible to know if these tests are conservative or over-sensitive in this case? Could a generalized linear model with gamma distribution be more appropriate? Thank you very much in advance for your responses.
Cheers,
Laurence
E-mail: [log in to unmask]
|