The influence of violations of assumptions on multilevel parameter estimates and their standard errors

By Cora J. M. Maas, Joop J. Hox. Computational Statistics & Data Analysis, 2004

410 citations recorded by the source corpus.

View Paper: https://www.semanticscholar.org/paper/76774b8b0218f29458f88e97b8c92be4a76070e8

How researchers see this paper

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

In Statistics and Probability

Citations in the corpus

On the Unnecessary Ubiquity of Hierarchical Linear Modeling (2017, background): “Furthermore, for the modest sample sizes often seen in psychology, power will be quite low and, the standard error estimates of variance components are highly sensitive to the deviations from normality (Maas & Hox, 2004).”
How Many Countries for Multilevel Modeling? A Comparison of Frequentist and Bayesian Approaches (2013, methodology): “Comprehensive examples are the studies conducted by Maas and Hox (2004b, 2005). For an overview of earlier studies, which use approximate estimation techniques, see Rodríguez and Goldman (1995) and Kreft (1996). A comprehensive study by Moineddin, Matheson, and Glazier (2007), oriented towards healt”
The Effect of Teacher Psychological and School Organizational and Leadership Factors on Teachers' Professional Learning in Dutch Schools (2009, background): “However, as we noted, the total number of schools was too small to obtain correct estimates of the school-level variance (Maas & Hox, 2004a, 2004b).”
Best-Practice Recommendations for Estimating Cross-Level Interaction Effects Using Multilevel Modeling (2013, methodology): “In addition to the simulation, Mathieu et al. (2012) conducted a power analysis based on articles published in Journal of Applied Psychology from 2000 to 2010 and found that power has been quite low.”
Attitudes towards vaccines and intention to vaccinate against COVID-19: Implications for public health communications (2020, neutral): “To account for any possible heteroscedasticity from our sampling method, we obtained robust standard errors in OLS and logistic regression models [21].”
The Impact of a Comprehensive Tier I Core Kindergarten Program on the Achievement of Students at Risk in Mathematics (2011, background): “…methods have also been found quite robust to violations of normality (Fitzmaurice et al., 2004), and several studies have found that nonnormal data lead to acceptable results in a variety of multilevel modeling scenarios (Donner & Klar, 1996; Maas & Hox, 2004a, 2004b; Murray et al., 2006).”
Cross-Level Influences of Empowering Leadership on Citizenship Behavior (2017, methodology): “…estimates and significance tests for multilevel and nonindependent data by estimating within-team and between-team variances and covariances separately and by using the robust standard errors for both within-group and between-group effects (Bliese, 2000; Maas & Hox, 2004; Raudenbush & Bryk, 2002).”
What We Know and Don’t Know About Corporate Social Responsibility (2012, methodology): “…variables and individual outcomes will lead to gross errors of prediction if a researcher uses statistical approaches, such as ordinary least squares regression, that are not designed to model data structures that include nonindependence due to clustering of entities (Maas & Hox, 2004).”
Analyzing Cross-Sectionally Clustered Data Using Generalized Estimating Equations (2021, background): “However, in reality, the assumptions about the distribution of the random effects in MLMs have a negligible impact on fixed effect estimates (Maas & Hox, 2004).”
Examining the Rule of Thumb of Not Using Multilevel Modeling: The “Design Effect Smaller Than Two” Rule (2015, methodology): “The numbers of clusters were 20, 30, 50, and 100, which is consistent with Maas and Hox (2004) and smaller than the ones used in B. O. Muthén and Satorra (1995).”
Quantifying individual variation in behaviour: mixed-effect modelling approaches. (2013, background): “Recommendations about optimal sample sizes vary substantially, depending on (i) what one aims to optimize (accuracy or power), (ii) the variance component of interest, and (iii) constraints imposed by the study system (Snijders & Bosker 1999; Maas & Hox 2004; © 2012 The Authors.”