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 10:24:29 -0400 Thanks, Vladimir, That means the random effects distribution of the parameter is perfectly matching to the radom effect distribution of the covariates? - when eta goes to zero after addition of covariates. Isn't there an assumption that the radom effect distriution is normal or lognormal? If so, does this mean the above possibility can never happen to parameters only with dichotomous covariates? - since dichotomous covariatres are in a quite different distribution type. Or it's also possible? - since the dichotomous covariates could be twisted to be a normal / lognormal distribution (I guess at least in the structural model, they are partly, if not completely, done this way since their effects are components of the radom effect distribution of the parameter - Am I wrong?). Thanks, Alan.