I would like to estimate the effect of a candidate's name on the proportion of negative preference votes she receives. My dependent variable "negative preference votes" measures the percentage of occasions a candidate is crossed off from a party ballot. The variable is bounded within a range [0,1]. I have used the GAMLSS package to fit a Beta-zero-one inflated distribution (BEINF). My observations are furthermore clustered in municipalities and candidates - but data structure is cross-nested. I would like to add cross-classified random effects to the model but I have not found any documentation on how to model that in GAMLSS.
As far as I am informed, there are two functions for fitting random effects within GAMLSS model, random() and re(). I have used the re() function and estimated the following model with two random effects, but I doubt that this specifies cross-classified random effects.
gamlss(prefvotes ~ candname, + re(random=~1|municipality) + re(random=~1|candidate), family=BEINF, data=nr_15)
Is there a way to specify cross-classified random effects in GAMLSS?
