2016
DOI: 10.1111/2041-210x.12623
|View full text |Cite
|
Sign up to set email alerts
|

Marginal or conditional regression models for correlated non‐normal data?

Abstract: 11. Correlated data are ubiquitous in ecological and evolutionary research, and appropriate statistical analysis requires that these correlations are taken into account. For regressions with correlated, non-normal outcomes, two main approaches are used: conditional and marginal modelling. The former leads to generalized linear mixed models (GLMMs), while the latter are estimated using generalized estimating equations (GEEs), or marginalized multilevel regression models. Differences, advantages and drawbacks of… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
1
1
1
1

Citation Types

0
39
0

Year Published

2018
2018
2024
2024

Publication Types

Select...
6
2

Relationship

2
6

Authors

Journals

citations
Cited by 36 publications
(39 citation statements)
references
References 81 publications
0
39
0
Order By: Relevance
“…Calculation of the marginal (R²m) and conditional (R²c) coefficient of determination was done using the MuMIn package (Barton, 2019). The R²m represents the variance in the dependent behavioral variable explained by the fixed effect of the respective CSD variable (across subjects), while the R²c reflects the total variance explained by the model's fixed and random effects, respectively (Muff et al, 2016). In a binary GLMM, the R 2 m is independent of sample size and dimensionless, which allows comparing fits across different datasets (Nakagawa and Schielzeth, 2013).…”
Section: Methodsmentioning
confidence: 99%
“…Calculation of the marginal (R²m) and conditional (R²c) coefficient of determination was done using the MuMIn package (Barton, 2019). The R²m represents the variance in the dependent behavioral variable explained by the fixed effect of the respective CSD variable (across subjects), while the R²c reflects the total variance explained by the model's fixed and random effects, respectively (Muff et al, 2016). In a binary GLMM, the R 2 m is independent of sample size and dimensionless, which allows comparing fits across different datasets (Nakagawa and Schielzeth, 2013).…”
Section: Methodsmentioning
confidence: 99%
“…that is, they cannot account for functional responses. Further, the slope estimator from a logistic model that omits random effects is a biased estimator of the mean slope in the population, a fact that has been discussed repeatedly in the statistical and ecological literature (e.g., Fieberg et al, 2009;Muff, Held, & Keller, 2016). Second, omitting individual-specific random slopes when they actually do vary between individuals induces too little uncertainty in the estimated parameters (e.g., Schielzeth & Forstmeier, 2009).…”
Section: The Importance Of Random Slopesmentioning
confidence: 99%
“…To address these challenges, several approaches to circumvent direct random-effects estimation have been proposed, such as the use of generalized estimating equations (GEEs, Craiu, Duchesne, & Fortin, 2008) or a two-step estimation approach (Craiu et al, 2011). GEEs, however, provide marginal parameter estimates that are analogous to those obtained from models without random effects, which are known to underestimate the true effect sizes experienced by individual animals (Fieberg et al, 2009;Lee & Nelder, 2004;Muff et al, 2016); thus, we do not generally recommend them for habitat-selection studies. The two-step approach is an efficient alternative that combines estimates of individual-specific regression parameters from standard ML methods for independent data with an expectationmaximization algorithm in conjunction with conditional restricted maximum likelihood (REML).…”
Section: Computational Challenges For Ssfsmentioning
confidence: 99%
“…Regression models that incorporate random effects offer a powerful approach to studying inter-individual variability and are frequently used to accommodate non-independent data in ecological studies (Fieberg et al, 2009;Muff et al, 2016). Gillies et al (2006) recommended using random intercepts to account for unequal sample sizes in habitatselection studies, and random slope coefficients (equivalently denoted as random coeffi-cients or random slopes) to account for individual-specific differences in habitat selection.…”
Section: Introductionmentioning
confidence: 99%
“…Given the challenges with fitting mixed conditional logistic regression models, it is not surprising that several approaches to circumvent direct random effects estimation have been proposed, such as the use of generalized estimating equations (GEEs, Craiu et al, 2008) or a two-step estimation approach (Craiu et al, 2011). GEEs, however, provide marginal parameter estimates that tend to underestimate the true effect sizes experienced by individual animals (Lee and Nelder, 2004;Fieberg et al, 2009;Muff et al, 2016); thus, we do not generally recommend them for habitat-selection studies.…”
mentioning
confidence: 99%