Page 1 of 1

Wald tests for random parameters

Posted: Tue Feb 16, 2010 1:58 pm
by Mark.McCann
Hi folks,

How can you calculate a Wald test by hand/calculator for more than one parameter? I'm not working in MLwiN and don't have the equivalent 'intervals and tests' window figured out.

Just in case it's different depending on which combination; I'm interested in testing the addition of random slopes, hence (I think) its the wald test for the variance of the coefficient, and the covariance of the coefficient & intercept.

Secondly, I'm working with logistic models, my documentation mentions 'approximate wald tests' for these parameters, and for the variance of the intercept I think I should be using a Chi squared test, how 'approximate' is the Wald test, and what in effect should that mean in terms of interpretability.

In other words, should I need much more convincing evidence to refute a null hypothesis, or should I be tenuous in my interpretation, whether accepting or refuting hypotheses?