Generalized linear models with clustered data: Fixed and random effects models

Broström, Göran; Holmberg, Henrik

doi:10.1016/j.csda.2011.06.011

Cited by 79 publications

(63 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As the cluster size increases, however, AGQ takes over and becomes most reliable in terms of bias and precision. Since the Laplace approximation is known to be precise only for normally distributed data or for non-normal data in large clusters [30], we observed an expected poor performance of this approximation in our settings [23]. PQL also exhibits an inferior performance for low icc's and a low outcome prevalence, while additionally revealing disconcerting performance issues for a withincluster measured predictor.…”

Section: Discussionmentioning

confidence: 78%

“…Focusing on the two aforementioned frameworks, current literature on the analysis of clustered binary outcomes reveals two major limitations: clusters of size two were either not considered [17][18][19][20], or they were, but limited to only one of both frameworks [21][22][23][24]. Here, we compare several estimation procedures within both GLMM-and SEM-frameworks for modeling this type of data, by considering the performance of relevant R-packages.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A Review of R-packages for Random-Intercept Probit Regression in Small Clusters

Josephy

Loeys

Rosseel

2016

Front. Appl. Math. Stat.

View full text Add to dashboard Cite

Generalized Linear Mixed Models (GLMMs) are widely used to model clustered categorical outcomes. To tackle the intractable integration over the random effects distributions, several approximation approaches have been developed for likelihood-based inference. As these seldom yield satisfactory results when analyzing binary outcomes from small clusters, estimation within the Structural Equation Modeling (SEM) framework is proposed as an alternative. We compare the performance of R-packages for random-intercept probit regression relying on: the Laplace approximation, adaptive Gaussian quadrature (AGQ), Penalized Quasi-Likelihood (PQL), an MCMC-implementation, and integrated nested Laplace approximation within the GLMM-framework, and a robust diagonally weighted least squares estimation within the SEM-framework. In terms of bias for the fixed and random effect estimators, SEM usually performs best for cluster size two, while AGQ prevails in terms of precision (mainly because of SEM's robust standard errors). As the cluster size increases, however, AGQ becomes the best choice for both bias and precision.

show abstract

Section: Discussionmentioning

confidence: 78%

Section: Introductionmentioning

confidence: 99%

A Review of R-packages for Random-Intercept Probit Regression in Small Clusters

Josephy

Loeys

Rosseel

2016

Front. Appl. Math. Stat.

View full text Add to dashboard Cite

show abstract

“…The Wald test was used to assess the significance. Statistical analyses were performed with R v. 2.15.1 [56] using the packages BiodiversityR [57], ape [58], lme4 [59] and glmmML [60].…”

Section: (D) Phylogenetic Analysismentioning

confidence: 99%

Effects of logging and recruitment on community phylogenetic structure in 32 permanent forest plots of Kampong Thom, Cambodia

Toyama

Kajisa

Tagane

et al. 2015

Phil. Trans. R. Soc. B

View full text Add to dashboard Cite

Ecological communities including tropical rainforest are rapidly changing under various disturbances caused by increasing human activities. Recently in Cambodia, illegal logging and clear-felling for agriculture have been increasing. Here, we study the effects of logging, mortality and recruitment of plot trees on phylogenetic community structure in 32 plots in Kampong Thom, Cambodia. Each plot was 0.25 ha; 28 plots were established in primary evergreen forests and four were established in secondary dry deciduous forests. Measurements were made in 1998, 2000, 2004 and 2010, and logging, recruitment and mortality of each tree were recorded. We estimated phylogeny using rbcL and matK gene sequences and quantified phylogenetic a and b diversity. Within communities, logging decreased phylogenetic diversity, and increased overall phylogenetic clustering and terminal phylogenetic evenness. Between communities, logging increased phylogenetic similarity between evergreen and deciduous plots. On the other hand, recruitment had opposite effects both within and between communities. The observed patterns can be explained by environmental homogenization under logging. Logging is biased to particular species and larger diameter at breast height, and forest patrol has been effective in decreasing logging.

show abstract

“…Computations were completed using the 'glmmML' package (Broström and Holmberg, 2012) in the statistical software R, version 2.14.1 (R Development Core Team, 2012).…”

Section: Data and Analysismentioning

confidence: 99%

Relationships between the Abundance of Abies sachalinensis Juveniles and Site Conditions in Selection Forests of Central Hokkaido, Japan

Owari

2013

FORMATH

View full text Add to dashboard Cite

Abstract:This case study analysis examines how overstory and topographic conditions impact regeneration of Abies sachalinensis (Abies) in a selection forest at the University of Tokyo Hokkaido Forest in Northern Japan. The analysis includes 1,382 inventory plots that were regularly surveyed between 1995 and 2004. The density of Abies juveniles and basal area of mature Abies overstory trees were calculated for each plot. Five topographic conditions-elevation, slope angle, curvature, solar radiation, and the topographic wetness index-were examined through spatial analyses using ArcGIS software with 10-m grid digital elevation models. A general linear mixed model analysis was performed to determine site factors affecting the abundance of Abies juveniles. Results indicated that basal area of mature overstory Abies, slope angle, curvature, and solar radiation had positive effects on the abundance of Abies juveniles, while elevation had a negative effect. These factors may be related to the distribution of dwarf bamboo (Sasa senanensis and S. kurilensis) and pathogenic fungus (Racodium therryanum or Herpotrichia juniperi), both of which are recognized as major inhibitors of natural regeneration in Hokkaido.

show abstract

Generalized linear models with clustered data: Fixed and random effects models

Cited by 79 publications

References 7 publications

A Review of R-packages for Random-Intercept Probit Regression in Small Clusters

A Review of R-packages for Random-Intercept Probit Regression in Small Clusters

Effects of logging and recruitment on community phylogenetic structure in 32 permanent forest plots of Kampong Thom, Cambodia

Relationships between the Abundance of Abies sachalinensis Juveniles and Site Conditions in Selection Forests of Central Hokkaido, Japan

Contact Info

Product

Resources

About