Dear Richard,
Singularities are a concern for most problems. The REML loglik, fitted
values are invariant, however estimates of fixed effects are not. I
would take care to not overfit, ie fit a singular model, unless you are
sure you know what you are doing. I also have some concern with
estimates of se's in the presence of singularities, though Arthur may
ally these fears
cheers
brian cullis
On Mon, 16 Mar 1998 richardm@calm.wa.gov.au wrote:
> Hi,
> I am interested /concerned about numbers of singularities occuring in
> different analyses. It seems that factors which affect the number of
> singularities include data balance and model specification.
>
> Can any one enlighten me as to how critical it is to eliminate
> singularities, or how many is too many?
> i.e is there a general guideline?
>
> thanks
>
> Richard
>
>
> ======================
> Richard Mazanec
> Dept of Conservation and Land Management,
> Science and Information Division,
> Brain Street,
> Manjimup. 6258
> Phone 08 9771 7985 Int. +61 8 9771 7985
> Fax 08 9777 1183 Int +61 8 9777 1183
>
...............................................................................
Brian Cullis Tel: 02 6938 1855
NSW Agriculture Fax: 02 6938 1809
Wagga Agricultural Institute email: brian.cullis@agric.nsw.gov.au
Pine Gully Rd
WAGGA WAGGA NSW 2650