Return to Revisions

If you want 200 copies of each of 5 levels in a polytomous variable in random order, then do this instead:

 x <- sample(rep(paste0('pict', 1:5), 200))

If you want to control for overall prevalence of a specific outcome, then you must choose which beta you will fudge. I usually do beta0.

 MM         <- model.matrix(~x)
 betas      <- rnorm(4)
 prevTarget <- 0.3
 prevDiff   <- function(beta0)  prevTarget - 
                                mean(binomial()$linkinv(MM%*%c(beta0, betas)))
beta0      <- uniroot(prevDiff, c(-100, 100))$root
 mean(binomial()$linkinv(MM%*%c(beta0, betas)))