Dear all,
I'm faced with the problem of modelling a highly multidimensional
spatially correlated response as a function of some covariates. I had
quite interesting initial results by building two separate spatial
regression models for the first two pc's of the multidimensional
response, though I'm obviously unsatisfied with the methodology. My idea
is that a theoretically better way to deal with this data structure
would be in terms of structural equation models, but I'm not very
familiar with those. Mainly I'm not really sure if standard softwares
(like SAS PROC CALIS or LISREL) can consider spatial correlation
structures as for example within mixed effects models. Is anybody aware
of such a kind of use of sem's? Does any other solution come to your
mind for the problem at hand?
Thank you in advance,
Alessio Pollice
--
Alessio Pollice
Dipartimento di Scienze Statistiche
Università delgi Studi di Bari
Via C. Rosalba 53
70124 Bari, ITALY
|