Displaying 1 result from an estimated 1 matches for "llrt".
Did you mean:
lldt
2009 Jan 07
0
fixed effect significance_NB mixed models_further pursuit
...-effects negative binomial
regressions.
From what I gather, there is no consensus on how best to go about
this; although there are two main lines of thought (at least as would
apply to someone in my case with limited statistical and programming
experience):
1) Log-likelihood ratio tests (LLRT), comparing the saturated model to
reduced models (wherein each predictor is dropped in turn); and
2) Wald tests, using parameter estimates and parameter standard errors
to calculate the Wald statistic.
I understand there are problems inherent to each approach. Following
the logic of D. Bat...