What distribution types should be specified for non-normal continuous data in GLMMs?

[tan1721]

Following is the histogram and Q-Q plot of deceleration data retrieved from driving simulator experiment. For this non- normally distributed data, I am using generalized linear mixed model to analyze the impact of gender and age on the deceleration rate. but the analysis is not working if I specify the distribution type as gamma, it is working only if I specify the data as normal. In that case, which distribution I should specify in GLMM? Does it become linear mixed model if the distribution is specified as normal in proc glimmix? 

[StatDave · Accepted solution]

You might want to see if the gamma distribution works reasonably well with the absolute value of your variable. See this note if the absolute value of the variable contains zeros.

[tan1721]

It worked.  I had to increase the iteration numbers as well.

Thanks!