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,
KreditriskmodelleringPublication year: 2007Abstract: 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