Variance Partitioning in Multilevel Logistic Models that Exhibit Overdispersion

Abstract

A common application of multilevel models is to apportion the variance in the response according to the different levels of the data. Whereas partitioning variances is straightforward in models with a continuous response variable with a normal error distribution at each level, the extension of this partitioning to models with binary responses or to proportions or counts is less obvious. We describe methodology due to Goldstein and co-workers for apportioning variance that is attributable to higher levels in multilevel binomial logistic models. This partitioning they referred to as the variance partition coefficient. We consider extending the variance partition coefficient concept to data sets when the response is a proportion and where the binomial assumption may not be appropriate owing to overdispersion in the response variable. Using the literacy data from the 1991 Indian census we estimate simple and complex variance partition coefficients at multiple levels of geography in models with significant overdispersion and thereby establish the relative importance of different geographic levels that influence educational disparities in India. Copyright 2005 Royal Statistical Society.

Bringing the individual back to small-area variation studies: A multilevel analysis of all-cause mortality in Andalusia, Spain

“…Browne et al, 2005;Goldstein et al, 2002;Li J et al, 2008;Merlo et al, 2006b;Snijders & Bokser, 1999), as follows:…”

“…It is necessary to quantify how much of the total individual variation is at the area level (W. J. Browne et al, 2005;Clarke P & Wheaton B, 2007;Larsen & Merlo, 2005;Merlo, 2003;Merlo et al, 2009), as we have done in the present study by calculating the VPCs. Furthermore, because of technical difficulties, ecological/spatial analyses only consider a few individual characteristics (e.g.…”

The Dynamic Risk of Heavy Episodic Drinking on Interpersonal Assault in Young Adolescence and Early Adulthood

“…However in binary response models there is no such equivalence. xii This was calculated using the latent variable approach (see Snijders and Bosker (2012) for a description and Browne et al (2005) for an analytical critique of this and other approaches). xiii It may at first sight seem confusing that we have a random effect of occasion and a fixed effect of sweep year since they appear conflated.…”

09:58: Test‐Retest Reliability of Drug‐Induced Sleep Endoscopy

“…By determining the variance in DISE findings attributable to different sources, it is possible to calculate results analogous to intraclass correlation coefficients for logistic models. 16 Four separate such testretest intraclass correlation coefficient analogs are generated to describe the test-retest reliability: correlation from Evaluation 1 to Evaluation 2 by the same reviewer and same exam (ICC Evaluation), correlation from blinded to unblinded reviewer for the same exam (ICC Reviewer), correlation from Exam 1 to Exam 2 for the same reviewer (ICC Exam), and correlation across exams and reviewers (ICC Reviewer-Exam). For all results, point estimates of the intraclass correlation analogs are reported with 95% confidence intervals.…”