I have a data set with an exponentially distributed outcome variable (it also includes values of zero). Is there a way to fit a linear mixed model to this data? And does anyone know how to do this in the lme4 package in R?
You can do this using the glmer function. The family should be Gamma. You will get the summary of the exponential model setting the argument dispersion=1 in the summary function (the gamma distribution with dispersion parameter fixed at 1 is the exponential distribution). If you don't fix the dispersion, the MLE is used.
Thanks! And would this be the best way also if the model is interpreted as an ANOVA-type model? I am looking to get reliable p-value estimates.
It can be difficult to pinpoint the null hypothesis in (G)LMMs, that's not always as straight-forward as in fixed-effect models.
https://www.ssc.wisc.edu/sscc/pubs/MM/MM_TestEffects.html
https://daijiang.name/en/2015/06/22/why-no-p-values-in-mixed-models/
- Badges
- Science topic
- Similar topics
- Geoscience
- Asia
- India