Hi, This question is regarding choice of R formula for my mixed-effects model. My response variable is a 3-level discrete variable. The independent variables comprise of both continuous and categorical variables. Furthermore, one of the categorical variables qualifies as a random effect. I am modeling it as a generalized linear model. Which formula in R would be appropriate in this case? I would have used *glmer*, but as far as I could figure out, there is no appropriate option in '*family' *to model a 3-level response variable. Thanks, Abhimanyu
Appropriate formula for my mixed-effects model
2 messages · Abhimanyu Sahai, Ben Bolker
Are your three levels ordered or unordered? If they are ordered, you can use clmm from the ordinal package. If they're unordered, you can use the MCMCglmm package. It is also possible (with considerably more work) to construct multinomial responses by appropriately combining binomial fits (e.g. taking your three levels and coding two separate responses e.g. ["level 1" vs "not level 1"] and ["level 2" vs "not level 2"] , e.g. see chapter 9.4 of Dobson and Barnett "introduction to Generalized Linear Models". There may be other options but I can't think of them off the top of my head.
On 17-03-20 04:54 PM, Abhimanyu Sahai wrote:
Hi, This question is regarding choice of R formula for my mixed-effects model. My response variable is a 3-level discrete variable. The independent variables comprise of both continuous and categorical variables. Furthermore, one of the categorical variables qualifies as a random effect. I am modeling it as a generalized linear model. Which formula in R would be appropriate in this case? I would have used *glmer*, but as far as I could figure out, there is no appropriate option in '*family' *to model a 3-level response variable. Thanks, Abhimanyu [[alternative HTML version deleted]]
_______________________________________________ R-sig-mixed-models at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models