Computanionally feasible estimation of the covariance structure in Generalized linear mixed models (GLMM)

Document identifier: oai:dalea.du.se:1793
Keyword: Monte-Carlo simulations, Large sample, Interdependence, Cluster errors, Kreditriskmodellering
Publication year: 2007
Abstract:

In this paper we discuss how a regression model, with a non-continuous response variable, that allows for dependency between observations should be estimated when observations are clustered and there are repeated measurements on the subjects. The cluster sizes are assumed to be large. We find that the conventional estimation technique suggested by the literature on Generalized Linear Mixed Models (GLMM) is slow and often fails due to non-convergence and lack of memory on standard PCs. We suggest to estimate the random effects as fixed effects by GLM and derive the covariance matrix from these estimates. A simulation study shows that our proposal is feasible in terms of Mean-Square Error and computation time. We recommend that our proposal be implemented in the software of GLMM techniques so that the estimation procedure can switch between the conventional technique and our proposal depending on the size of the clusters.

Authors

Kenneth Carling

Högskolan Dalarna; Statistik
Other publications >>

M.M. Alam

Other publications >>

Record metadata

Click to view metadata

	
    header: 
        identifier: oai:dalea.du.se:1793
        datestamp: 2021-04-15T12:05:48Z
        setSpec: SwePub-du
    metadata: 
        mods: 
            @attributes: 
                version: 3.7
            recordInfo: 
                recordContentSource: du
                recordCreationDate: 2006-02-24
            identifier: http://urn.kb.se/resolve?urn=urn:nbn:se:du-1793
            titleInfo: 
                @attributes: 
                    lang: eng
                title: Computanionally feasible estimation of the covariance structure in Generalized linear mixed models (GLMM)
            abstract: 
In this paper we discuss how a regression model with a non-continuous response variable\nthat allows for dependency between observations should be estimated when observations are\nclustered and there are repeated measurements on the subjects. The cluster sizes are assumed\nto be large. We find that the conventional estimation technique suggested by the literature\non Generalized Linear Mixed Models (GLMM) is slow and often fails due to non-convergence\nand lack of memory on standard PCs. We suggest to estimate the random effects as fixed\neffects by GLM and derive the covariance matrix from these estimates. A simulation study\nshows that our proposal is feasible in terms of Mean-Square Error and computation time. We\nrecommend that our proposal be implemented in the software of GLMM techniques so that\nthe estimation procedure can switch between the conventional technique and our proposal\ndepending on the size of the clusters.

            subject: 
                    @attributes: 
                        lang: eng
                    topic: Monte-Carlo simulations
                    @attributes: 
                        lang: eng
                    topic: large sample
                    @attributes: 
                        lang: eng
                    topic: interdependence
                    @attributes: 
                        lang: eng
                    topic: cluster errors
                    @attributes: 
                        lang: swe
                        authority: du
                    topic: Kreditriskmodellering
                    genre: Research subject
            language: 
                languageTerm: eng
            genre: 
                publication/report
                vet
            note: 
                Published
                2
            name: 
                    @attributes: 
                        type: personal
                        authority: du
                    namePart: 
                        Carling
                        Kenneth
                    role: 
                        roleTerm: aut
                    affiliation: 
                        Högskolan Dalarna
                        Statistik
                    nameIdentifier: 
                        kca
                        0000-0003-2317-9157
                    @attributes: 
                        type: personal
                    namePart: 
                        Alam
                        M.M.
                    role: 
                        roleTerm: aut
            originInfo: 
                dateIssued: 2007
                publisher: ESA Högskolan i Örebro
                place: 
                    placeTerm: Örebro
            relatedItem: 
                @attributes: 
                    type: series
                titleInfo: 
                    title: Working paper series
                    partNumber: 2007:14
                identifier: 1403-0586
            physicalDescription: 
                form: print
            typeOfResource: text