Cargando…
The Influence of Sample Size on Parameter Estimates in Three-Level Random-Effects Models
In educational psychology, observational units are oftentimes nested within superordinate groups. Researchers need to account for hierarchy in the data by means of multilevel modeling, but especially in three-level longitudinal models, it is often unclear which sample size is necessary for reliable...
Autores principales: | , |
---|---|
Formato: | Online Artículo Texto |
Lenguaje: | English |
Publicado: |
Frontiers Media S.A.
2019
|
Materias: | |
Acceso en línea: | https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6536630/ https://www.ncbi.nlm.nih.gov/pubmed/31164847 http://dx.doi.org/10.3389/fpsyg.2019.01067 |
_version_ | 1783421795349561344 |
---|---|
author | Kerkhoff, Denise Nussbeck, Fridtjof W. |
author_facet | Kerkhoff, Denise Nussbeck, Fridtjof W. |
author_sort | Kerkhoff, Denise |
collection | PubMed |
description | In educational psychology, observational units are oftentimes nested within superordinate groups. Researchers need to account for hierarchy in the data by means of multilevel modeling, but especially in three-level longitudinal models, it is often unclear which sample size is necessary for reliable parameter estimation. To address this question, we generated a population dataset based on a study in the field of educational psychology, consisting of 3000 classrooms (level-3) with 55000 students (level-2) measured at 5 occasions (level-1), including predictors on each level and interaction effects. Drawing from this data, we realized 1000 random samples each for various sample and missing value conditions and compared analysis results with the true population parameters. We found that sampling at least 15 level-2 units each in 35 level-3 units results in unbiased fixed effects estimates, whereas higher-level random effects variance estimates require larger samples. Overall, increasing the level-2 sample size most strongly improves estimation soundness. We further discuss how data characteristics influence parameter estimation and provide specific sample size recommendations. |
format | Online Article Text |
id | pubmed-6536630 |
institution | National Center for Biotechnology Information |
language | English |
publishDate | 2019 |
publisher | Frontiers Media S.A. |
record_format | MEDLINE/PubMed |
spelling | pubmed-65366302019-06-04 The Influence of Sample Size on Parameter Estimates in Three-Level Random-Effects Models Kerkhoff, Denise Nussbeck, Fridtjof W. Front Psychol Psychology In educational psychology, observational units are oftentimes nested within superordinate groups. Researchers need to account for hierarchy in the data by means of multilevel modeling, but especially in three-level longitudinal models, it is often unclear which sample size is necessary for reliable parameter estimation. To address this question, we generated a population dataset based on a study in the field of educational psychology, consisting of 3000 classrooms (level-3) with 55000 students (level-2) measured at 5 occasions (level-1), including predictors on each level and interaction effects. Drawing from this data, we realized 1000 random samples each for various sample and missing value conditions and compared analysis results with the true population parameters. We found that sampling at least 15 level-2 units each in 35 level-3 units results in unbiased fixed effects estimates, whereas higher-level random effects variance estimates require larger samples. Overall, increasing the level-2 sample size most strongly improves estimation soundness. We further discuss how data characteristics influence parameter estimation and provide specific sample size recommendations. Frontiers Media S.A. 2019-05-21 /pmc/articles/PMC6536630/ /pubmed/31164847 http://dx.doi.org/10.3389/fpsyg.2019.01067 Text en Copyright © 2019 Kerkhoff and Nussbeck. http://creativecommons.org/licenses/by/4.0/ This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms. |
spellingShingle | Psychology Kerkhoff, Denise Nussbeck, Fridtjof W. The Influence of Sample Size on Parameter Estimates in Three-Level Random-Effects Models |
title | The Influence of Sample Size on Parameter Estimates in Three-Level Random-Effects Models |
title_full | The Influence of Sample Size on Parameter Estimates in Three-Level Random-Effects Models |
title_fullStr | The Influence of Sample Size on Parameter Estimates in Three-Level Random-Effects Models |
title_full_unstemmed | The Influence of Sample Size on Parameter Estimates in Three-Level Random-Effects Models |
title_short | The Influence of Sample Size on Parameter Estimates in Three-Level Random-Effects Models |
title_sort | influence of sample size on parameter estimates in three-level random-effects models |
topic | Psychology |
url | https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6536630/ https://www.ncbi.nlm.nih.gov/pubmed/31164847 http://dx.doi.org/10.3389/fpsyg.2019.01067 |
work_keys_str_mv | AT kerkhoffdenise theinfluenceofsamplesizeonparameterestimatesinthreelevelrandomeffectsmodels AT nussbeckfridtjofw theinfluenceofsamplesizeonparameterestimatesinthreelevelrandomeffectsmodels AT kerkhoffdenise influenceofsamplesizeonparameterestimatesinthreelevelrandomeffectsmodels AT nussbeckfridtjofw influenceofsamplesizeonparameterestimatesinthreelevelrandomeffectsmodels |