compute denominator degrees-of-freedom approximations

dof_KR uses an adaptation of the machinery from the pbkrtest package to compute the Kenward-Roger approximation of the 'denominator degrees of freedom' for each fixed-effect coefficient in the conditional model; dof_satt does the same for Satterthwaite approximations

dof_KR(model)

dof_satt(model, L = diag(length(fixef(model)$cond)))

Arguments

model

a fitted glmmTMB object

L

a by default, equal to an identity matrix (i.e., ddfs are returned for each fixed-effect parameter

Value

a named vector of ddf for each conditional fixed-effect parameter; dof_KR includes attributes 'vcov' (Kenward-Roger adjusted covariance matrix) and 'se' (the corresponding standard errors)

Details

Kenward-Roger adjustments should not be used for models fitted with ML rather than REML; the theory is only well understood, and the model is only tested, for LMMs (family = "gaussian"). Use at your own risk for GLMMs!