Are multilevel techniques necessary?: An overview, including simulation studies

By I. Kreft. 2005

345 citations recorded by the source corpus.

View Paper: https://www.semanticscholar.org/paper/7187167f53c10eb74d363dac73d671414fed8683

How researchers see this paper

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

Citations in the corpus

Differentiation in leader-member exchange: A hierarchical linear modeling approach (2010, background): “Kreft (1996) suggested that, in general, large sample sizes are required: a sample size of 30 groups containing 30 individuals eachmay be needed for a sufficient power (.90) to detect cross-level interactions.”
How Many Countries for Multilevel Modeling? A Comparison of Frequentist and Bayesian Approaches (2013, methodology): “For an overview of earlier studies, which use approximate estimation techniques, see Rodríguez and Goldman (1995) and Kreft (1996).”
Modeling Clustered Data with Very Few Clusters (2016, background): “Very broadly, the 30-cluster / 30-unit recommendation for minimum sample size with clustered data that is often attributed to Kreft (1996) still permeates in much applied literature but is quickly being rendered obsolete, outdated, and inaccurate as methodological advances continue to burgeon.”
Sufficient Sample Sizes for Multilevel Modeling (2005, background): “In practice, 50 groups is a frequently occurring number in organizational and school research, and 30 is the smallest acceptable number according to Kreft and De Leeuw (1998). Similarly, the group sizes are chosen so that the highest number should be sufficient.”
On the Unnecessary Ubiquity of Hierarchical Linear Modeling (2017, methodology): “…HLM. Recommendations for how large is “large enough” vary across and within each method and common suggestions for HLM are 30 clusters of size 30 (Kreft, 1996), at least 20 clusters (Snijders & Bosker, 2011), or 50 clusters of size 20 for cross-level interactions or 100 clusters of size 10 for…”
Academic Achievement in the First Year of College: Evidence of the Pervasive Effects of the High School Context (2010, methodology): “particular study and we chose to focus on the student-level model using a blocked linear regression technique (Heinrich and Lynn 2001; Kreft 1996).”
The influence of work‐family culture and workplace relationships on work interference with family: a multilevel model (2008, background): “sample size should be at least 30 (Hox, 2002; Kreft, 1996; Maas & Hox, 2004).”
The Effect of Small Sample Size on Two-Level Model Estimates: A Review and Illustration (2014, background): “…longitudinally clustered data) with a cluster size of 30 (microunit, level-1 sample size, or repeated measures for longitudinally clustered data) in Kreft (1996), a minimum of 20 clusters (Snijders and Bosker 2012), or 50 clusters with a cluster size of 20 for cross-level interactions or 100…”
Teacher-student relationship and mathematical problem-solving ability: mediating roles of self-efficacy and mathematical anxiety (2020, result): “This ICC value was much smaller than 10%, the value found in many other school performance studies, and thus suggested the multilevel analyses approach might not be warranted (Hedges & Hedberg, 2007; Kreft, 1996; Kreft & de Leeuw, 1998).”
EDA for HLM: Visualization when Probabilistic Inference Fails (2005, background): “9For much more discussion of the benefits, weaknesses, details, and implementation of multilevel models see Steenbergen and Jones (2002); Snijders and Bosker (1999); Kreft and Leeuw (1998); Longford (1993); Goldstein (1999); Pinheiro and Bates (2000); Raudenbush and Bryk (2002); Singer and Willett (”