Implementation and performance issues in the Bayesian and likelihood fitting of multilevel models

By William J. Browne, Denise Draper. 2000

215 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Statistics and Probability

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Statistics and Probability

Citations in the corpus

A Bayesian model for repeated measures zero-inflated count data with application to outpatient psychiatric service use (2010, background): “Some Statistical Modelling 2010; 10(4): 421–439 authors have noted that I-W priors can yield biased and imprecise model estimates, particularly if the number of clustering units is small (Browne and Draper, 2000; Lambert et al., 2005).”
Regression Analysis of Country Effects Using Multilevel Data: A Cautionary Tale (2013, methodology): “Simulations by Browne and Draper (2000) and Pinheiro and Chao (2006) re-used the three-level data structure employed by Rodriguez and Goldman (1995) with relatively small NC.”
How and when does complex reasoning occur? Empirically driven development of a learning progression focused on complex reasoning about biodiversity (2009, background): “Although the number of groups at level 2 is small (n¼ 22), prior research has concluded that maximum likelihood in multilevel models with a small number of groups still provides unbiased estimates offixed effects such as our treatment1 (Browne & Draper, 2000; Van Der Leeden et al., 1997).”
Examining the Rule of Thumb of Not Using Multilevel Modeling: The “Design Effect Smaller Than Two” Rule (2015, result): “This was consistent with results in previous studies (e.g., Browne & Draper, 2000; Maas & Hox, 2005) that standard errors of fixed effects were relatively unbiased only when number of clusters was at least 30.”
The influence of violations of assumptions on multilevel parameter estimates and their standard errors (2004, background): “Browne and Draper (2000) report similar results.”
Bayesian and likelihood-based methods in multilevel modeling 1 A comparison of Bayesian and likelihood-based methods for fitting multilevel models (2006, methodology): “The fixed effects and residuals may also be block-updated using multivariate normal proposal distributions; Browne and Draper (2000) describes comparisons between these two Metropolis alternatives and documents the pronounced Monte Carlo efficiency advantage of the hybrid Metropolis-Gibbs approach…”
Leader-member exchange and affective organizational commitment: the contribution of supervisor's organizational embodiment. (2010, background): “…that with hierarchical linear modeling the absolute number of Level 2 units (i.e., the supervisors in the current study) is more important than the ratio of Level 2 to Level 1 units for the power and accuracy of fixed estimates (e.g., Bassiri, 1988; Browne and Draper, 2000; Kim, 1990; Mok, 1995).”
Modeling Heterogeneous Variance–Covariance Components in Two-Level Models (2014, background): “See Browne and Draper (2000) and Seltzer et al. (1996) for further discussion regarding sensitivity of model results to choice of priors.”
Multilevel modelling of medical data (2002, methodology): “A detailed comparison of Bayesian and likelihood procedures for �tting multilevel models is given in Browne and Draper [ 40 ].”
Estimating Statistical Power and Required Sample Sizes for Organizational Research Using Multilevel Modeling (2009, background): “Bassiri (1988), Brown and Daper (2000), Kim (1990), and Mok (1995) found that for fixed effects, increasing the sample size at Level 2 had a greater impact on increasing power than increasing the Level 1 sample size at multiple values of the intraclass correlation.”
An Introduction to MCMC for Machine Learning (2004, methodology): “A similar method that guarantees a particular acceptance rate is discussed in Browne and Draper (2000) .”
Sample sizes for multilevel modeling (2002, unknown): “However, their simulation results are comparable when only those conditions are considered that are similar to the conditions in our simulations and those of Browne and Draper (2000).”
Multilevel Modelling of Country Effects: A Cautionary Tale (2016, background): “See Austin (2010), Browne and Draper (2000, 2006), and Stegmueller (2013).”
The Impact of Sample Size and Other Factors When Estimating Multilevel Logistic Models (2016, background): “Level-2 sample size guidelines for linear multilevel models range from 6 to 100 level-2 units depending on whether interest is on fixed effects or variance components (Browne & Draper, 2000; Kreft & De Leeuw, 1998; Maas & Hox, 2004, 2005; Van Der Leeden, Busing, & Meijer, 1997).”