Denominator degrees of freedom
requests computation of the approximate
denominator degrees of freedom according to Kenward and Roger (1997)
for the significance testing of fixed
effects terms in the dense part of the linear mixed model. There are
three options for n:
n = -1 suppresses computation,
n = 1 and n = 2
compute the denominator d.f. using numerical and algebraic
methods respectively. If n is omitted then n = 2
is assumed, with
the proviso that n = -1 will be the default for 'large' jobs.
Calculation of the denominator degrees of freedom is computationally expensive.
Numerical derivatives require
an extra evaluation of the mixed model equations for every variance parameter.
require a large dense matrix, potentially of order number of equations
plus number of records and is not available when
is 1 or for multivariate analysis.
Return to start