Re: Help: Non-positive semi-definite message
From: Alan Xiao <Alan.Xiao@cognigencorp.com>
Subject: Re: Help: Non-positive semi-definite message
Date: Thu, 16 Aug 2001 11:07:42 -0400
I think I mixed the covariate's random effect distribution with the
covariate distribution.
Aren't their values measured and there are some kind of radom effects,
just like concentration records? - where residual variability is
(partly) defined. OK, we just ignore that by assumption - after all,
we can not cover every thing. However, covariates have their own
distributions. if a parameter can be expressed as:
P = theta0 + theta1*cov1 + theta2*cov2 + eta
then
var(P) = theta1**2*var(cov1) + theta2**2*var(cov2) + var(eta)
-assuming there is no random effect distribution on thetas.
now if cov1, cov2 and eta have different distribution types - assuming,
cov1=AGE, cov2=GENDER and we already assumed eta is in a normal
distribution - are they "normalized" to the normal distribution?
Thanks,
ALan.