[R] remove fixed effect in predict.merMod

[Alexander Shenkin]

Hello all,

I've run a model, and now would like to extract residuals.  However, I 
have set sum-to-zero contrasts for the categorical fixed effect 
(contr.sum).  Because I am interested in looking at the variation in the 
residuals associated with that fixed effect (along with other levels), I 
need to calculate residuals setting that fixed effect to zero.  Any 
thoughts on how to do this?

thanks,
allie



contr.sum.keepnames <- function(...) {
     # make deviation contrasts that don't lose the names of the factors 
in the model results
     # from 
https://stackoverflow.com/questions/10808853/why-does-changing-contrast-type-change-row-labels-in-r-lm-summary
     conS <- contr.sum(...)
     colnames(conS) = rownames(conS)[-length(rownames(conS))]
     conS
}

test_df = data.frame(site = rep(LETTERS[1:10], 10), resp = runif(100), 
pred = runif(100))

contrasts(test_df$site) = contr.sum.keepnames(levels(test_df$site))

mod = lmer(resp ~ (1 + pred|site) + pred, data = test_df)

residuals = test_df$resp - predict(mod, REform=NA) # I would like the 
site effect here to be set to zero