Sufficient Sample Sizes for Multilevel Modeling

By Cora J. M. Maas, Joop J. Hox. 2005

3,797 citations recorded by the source corpus.

Subfields: Genetics, Statistics and Probability

View Paper: https://www.semanticscholar.org/paper/7e51cac4bc694a1a3c07720fb33cee9ce37c4ba4

How researchers see this paper

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

In Genetics

In Statistics and Probability

Citations in the corpus

Preaching to different choirs: How to motivate dismissive, uncommitted, and alarmed audiences to adapt to climate change? (2016, background): “Maas and Hox (2005) ran a series of simulations in which they varied Level 1 and Level 2 sample sizes.”
Private self-consciousness in daily life: Relationships between rumination and reflection and well-being, and meaning in daily life (2017, positive): “A series of simulation analyses reported by Maas and Hox (2005) suggested that this sample provided adequate power to estimate level-1 covariances, the focus of our analyses.”
Direct and Extended Vocabulary Instruction in Kindergarten: Investigating Transfer Effects (2010, methodology): “In their simulation study, Maas and Hox (2005) found that multilevel models with 10 Level 2 units produced biased standard error estimates for the regression coefficients and variance components, as well as Level 2 variance components that were overestimated by approximately 25%.”
The Influence of Demographic Risk Factors on Children's Behavioral Regulation in Prekindergarten and Kindergarten (2011, methodology): “Because the ICC was relatively high and significant, and the Level 2 sample size was sufficient to obtain unbiased estimates at all time points, we conducted all additional analyses with multilevel models (Maas & Hox, 2005; Raudenbush & Bryk, 2002).”
The Role of Arts Participation in Students' Academic and Nonacademic Outcomes: A Longitudinal Study of School, Home, and Community Factors. (2013, background): “A review by Marsh and Kleitman (2002) found the same and cited this as support for the identification/ commitment hypothesis in which context-specific activity impacts one’s identification with and commitment to outcomes in that context.”
Sample Size Limits for Estimating Upper Level Mediation Models Using Multilevel SEM (2013, methodology): “Several researchers (Hox & Maas, 2001; Kreft & de Leeuw, 1998; Maas & Hox, 2005; Van der Leeden, Busing, & Meijer, 1997) have made recommendations about the optimal minimal number of clusters, G, to use when estimating MMs. High proportions of nonconvergent cases, and biased parameter and standard…”
An Investigation of the Alignment Method With Polytomous Indicators Under Conditions of Partial Measurement Invariance (2018, methodology): “…for testing for equivalence across many groups (De Jong, Steenkamp, & Fox, 2007; Muthén & Asparouhov, 2013b), these models rely on unbiased estimates of the Level 2 variances, requiring upward of 50 groups (Maas & Hox, 2005), a sample size that might not be attainable for many researchers.”
State Authenticity in Everyday Life (2016, background): “This sample size is more than adequate to yield accurate estimates of the statistics that we report in Table 1 (Maas & Hox, 2005).”
Alternative Methods for Assessing Mediation in Multilevel Data: The Advantages of Multilevel SEM (2011, methodology): “Assuming that researchers adhere to common sample size recommendations for MLM (Hox & Maas, 2001; Maas & Hox, 2005; Snijders & Bosker, 1999) when using MSEM (discussed later), these problems are minimized.”
An Introduction to Multilevel Modeling for Social and Personality Psychology (2008, background): “As discussed by Nezlek (2001) and Raudenbush and Bryk (2002), choices about centering should reflect substantive issues – what does the analyst want the intercept to represent and why? Blanket recommendations to center one way or another should be ignored because no single rule can cover all cases.”
Perceived coach-autonomy support, basic need satisfaction and the well- and ill-being of elite youth soccer players: A longitudinal investigation (2012, methodology): “The current sample size is in accordance with simulation data that recommend fifty or more units at level two in a multi-level model (Maas & Hox, 2005).”
A multilevel study on servant leadership, job boredom and job crafting (2018, background): “Third, the number of Level 2 units (i.e. 47 teams) was small for detecting both slope variance across Level 2 units and cross-level interaction effects, which is a common problem in studies assessing cross-level interactions (Mathieu et al., 2012; Scherbaum and Ferreter, 2009; Maas and Hox, 2005).”