If there are no positive observations for a contrast, it is generally not estimible. However, for the purposes of testing we can replace it with the least favorable value with respect to the contrasts that are defined.
impute(object, groupby)
object | Output of predict |
---|---|
groupby | Variables (column names in predict) to group by for imputation (facets of the plot) |
data.table
##See stat_ell example(stat_ell)#> Warning: no help found for ‘stat_ell’