Limitations of ordinary least squares models in analyzing repeated measures data

To a) introduce and present the advantages of linear mixed models using generalized least squares (GLS) when analyzing repeated measures data; and b) show how model misspecification and an inappropriate analysis using repeated measures ANOVA with ordinary least squares (OLS) methodology can negative...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Medicine and science in sports and exercise 2004-12, Vol.36 (12), p.2144-2148
Hauptverfasser: UGRINOWITSCH, Carlos, FELLINGHAM, Gilbert W, RICARD, Mark D
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page 2148
container_issue 12
container_start_page 2144
container_title Medicine and science in sports and exercise
container_volume 36
creator UGRINOWITSCH, Carlos
FELLINGHAM, Gilbert W
RICARD, Mark D
description To a) introduce and present the advantages of linear mixed models using generalized least squares (GLS) when analyzing repeated measures data; and b) show how model misspecification and an inappropriate analysis using repeated measures ANOVA with ordinary least squares (OLS) methodology can negatively impact the probability of occurrence of Type I error. The effects of three strength-training groups were simulated. Strength gains had two slope conditions: null (no gain), and moderate (moderate gain). Ten subjects were hypothetically measured at five time points, and the correlation between measurements within a subject was modeled as compound symmetric (CS), autoregressive lag 1 (AR(1)), and random coefficients (RC). A thousand data sets were generated for each correlation structure. Then, each was analyzed four times--once using OLS, and three times using GLS, assuming the following variance/covariance structures: CS, AR(1), and RC. OLS produced substantially inflated probabilities of Type I errors when the variance/covariance structure of the data set was not CS. The RC model was less affected by the actual variance/covariance structure of the data set, and gave good estimates across all conditions. Using OLS to analyze repeated measures data is inappropriate when the covariance structure is not known to be CS. Random coefficients growth curve models may be useful when the variance/covariance structure of the data set is unknown.
doi_str_mv 10.1249/01.mss.0000147580.40591.75
format Article
fullrecord <record><control><sourceid>proquest_cross</sourceid><recordid>TN_cdi_proquest_miscellaneous_67134290</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>17485675</sourcerecordid><originalsourceid>FETCH-LOGICAL-c496t-13faad15032086dd9d84342e11a182486ac65be13be4611aff7ab3311589a34c3</originalsourceid><addsrcrecordid>eNqFkE1r3DAQhkVJaDZp_0IRgfRmR2NJltVbCWla2JLAtmcxa8lFwR8bjX1Ifn20ycIeO5eB4XlnhoexSxAlVMpeCygHolLkAmV0I0oltIXS6A9sBVqKQkjQJ2wlwOrCgoQzdk70mHkjJXxkZ6C1EaCrFXtYxyHOOMdpJD51fEo-jpieeR-QZk5PC6ZAfJh86InHkeOI_fNLHP_xFHYB5-D5kNFlT3mc8RM77bCn8PnQL9jfH7d_bn4W6_u7Xzff10WrbD0XIDtED1rISjS199Y3SqoqACA0lWpqbGu9DSC3QdV52HUGt_l50I1FqVp5wb6-792l6WkJNLshUhv6HscwLeRqA3mfFf8FwahG10Zn8Ns72KaJKIXO7VIcsgsHwu3FOwHu92bjjuLdm3j3Fv5yuLJsh-CP0YPpDFwdAKQW-y7h2EY6crWUwkojXwHs14xD</addsrcrecordid><sourcetype>Aggregation Database</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>17485675</pqid></control><display><type>article</type><title>Limitations of ordinary least squares models in analyzing repeated measures data</title><source>MEDLINE</source><source>Journals@Ovid LWW Legacy Archive</source><source>Journals@Ovid Complete</source><creator>UGRINOWITSCH, Carlos ; FELLINGHAM, Gilbert W ; RICARD, Mark D</creator><creatorcontrib>UGRINOWITSCH, Carlos ; FELLINGHAM, Gilbert W ; RICARD, Mark D</creatorcontrib><description>To a) introduce and present the advantages of linear mixed models using generalized least squares (GLS) when analyzing repeated measures data; and b) show how model misspecification and an inappropriate analysis using repeated measures ANOVA with ordinary least squares (OLS) methodology can negatively impact the probability of occurrence of Type I error. The effects of three strength-training groups were simulated. Strength gains had two slope conditions: null (no gain), and moderate (moderate gain). Ten subjects were hypothetically measured at five time points, and the correlation between measurements within a subject was modeled as compound symmetric (CS), autoregressive lag 1 (AR(1)), and random coefficients (RC). A thousand data sets were generated for each correlation structure. Then, each was analyzed four times--once using OLS, and three times using GLS, assuming the following variance/covariance structures: CS, AR(1), and RC. OLS produced substantially inflated probabilities of Type I errors when the variance/covariance structure of the data set was not CS. The RC model was less affected by the actual variance/covariance structure of the data set, and gave good estimates across all conditions. Using OLS to analyze repeated measures data is inappropriate when the covariance structure is not known to be CS. Random coefficients growth curve models may be useful when the variance/covariance structure of the data set is unknown.</description><identifier>ISSN: 0195-9131</identifier><identifier>EISSN: 1530-0315</identifier><identifier>DOI: 10.1249/01.mss.0000147580.40591.75</identifier><identifier>PMID: 15570152</identifier><identifier>CODEN: MSPEDA</identifier><language>eng</language><publisher>Hagerstown, MD: Lippincott Williams &amp; Wilkins</publisher><subject>Biological and medical sciences ; Endpoint Determination ; Exercise - physiology ; Fundamental and applied biological sciences. Psychology ; Humans ; Least-Squares Analysis ; Models, Theoretical ; Reproducibility of Results ; Space life sciences ; Task Performance and Analysis ; Treatment Outcome ; Vertebrates: body movement. Posture. Locomotion. Flight. Swimming. Physical exercise. Rest. Sports ; Weight Lifting</subject><ispartof>Medicine and science in sports and exercise, 2004-12, Vol.36 (12), p.2144-2148</ispartof><rights>2005 INIST-CNRS</rights><lds50>peer_reviewed</lds50><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed><citedby>FETCH-LOGICAL-c496t-13faad15032086dd9d84342e11a182486ac65be13be4611aff7ab3311589a34c3</citedby><cites>FETCH-LOGICAL-c496t-13faad15032086dd9d84342e11a182486ac65be13be4611aff7ab3311589a34c3</cites></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>314,780,784,27924,27925</link.rule.ids><backlink>$$Uhttp://pascal-francis.inist.fr/vibad/index.php?action=getRecordDetail&amp;idt=16330937$$DView record in Pascal Francis$$Hfree_for_read</backlink><backlink>$$Uhttps://www.ncbi.nlm.nih.gov/pubmed/15570152$$D View this record in MEDLINE/PubMed$$Hfree_for_read</backlink></links><search><creatorcontrib>UGRINOWITSCH, Carlos</creatorcontrib><creatorcontrib>FELLINGHAM, Gilbert W</creatorcontrib><creatorcontrib>RICARD, Mark D</creatorcontrib><title>Limitations of ordinary least squares models in analyzing repeated measures data</title><title>Medicine and science in sports and exercise</title><addtitle>Med Sci Sports Exerc</addtitle><description>To a) introduce and present the advantages of linear mixed models using generalized least squares (GLS) when analyzing repeated measures data; and b) show how model misspecification and an inappropriate analysis using repeated measures ANOVA with ordinary least squares (OLS) methodology can negatively impact the probability of occurrence of Type I error. The effects of three strength-training groups were simulated. Strength gains had two slope conditions: null (no gain), and moderate (moderate gain). Ten subjects were hypothetically measured at five time points, and the correlation between measurements within a subject was modeled as compound symmetric (CS), autoregressive lag 1 (AR(1)), and random coefficients (RC). A thousand data sets were generated for each correlation structure. Then, each was analyzed four times--once using OLS, and three times using GLS, assuming the following variance/covariance structures: CS, AR(1), and RC. OLS produced substantially inflated probabilities of Type I errors when the variance/covariance structure of the data set was not CS. The RC model was less affected by the actual variance/covariance structure of the data set, and gave good estimates across all conditions. Using OLS to analyze repeated measures data is inappropriate when the covariance structure is not known to be CS. Random coefficients growth curve models may be useful when the variance/covariance structure of the data set is unknown.</description><subject>Biological and medical sciences</subject><subject>Endpoint Determination</subject><subject>Exercise - physiology</subject><subject>Fundamental and applied biological sciences. Psychology</subject><subject>Humans</subject><subject>Least-Squares Analysis</subject><subject>Models, Theoretical</subject><subject>Reproducibility of Results</subject><subject>Space life sciences</subject><subject>Task Performance and Analysis</subject><subject>Treatment Outcome</subject><subject>Vertebrates: body movement. Posture. Locomotion. Flight. Swimming. Physical exercise. Rest. Sports</subject><subject>Weight Lifting</subject><issn>0195-9131</issn><issn>1530-0315</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2004</creationdate><recordtype>article</recordtype><sourceid>EIF</sourceid><recordid>eNqFkE1r3DAQhkVJaDZp_0IRgfRmR2NJltVbCWla2JLAtmcxa8lFwR8bjX1Ifn20ycIeO5eB4XlnhoexSxAlVMpeCygHolLkAmV0I0oltIXS6A9sBVqKQkjQJ2wlwOrCgoQzdk70mHkjJXxkZ6C1EaCrFXtYxyHOOMdpJD51fEo-jpieeR-QZk5PC6ZAfJh86InHkeOI_fNLHP_xFHYB5-D5kNFlT3mc8RM77bCn8PnQL9jfH7d_bn4W6_u7Xzff10WrbD0XIDtED1rISjS199Y3SqoqACA0lWpqbGu9DSC3QdV52HUGt_l50I1FqVp5wb6-792l6WkJNLshUhv6HscwLeRqA3mfFf8FwahG10Zn8Ns72KaJKIXO7VIcsgsHwu3FOwHu92bjjuLdm3j3Fv5yuLJsh-CP0YPpDFwdAKQW-y7h2EY6crWUwkojXwHs14xD</recordid><startdate>20041201</startdate><enddate>20041201</enddate><creator>UGRINOWITSCH, Carlos</creator><creator>FELLINGHAM, Gilbert W</creator><creator>RICARD, Mark D</creator><general>Lippincott Williams &amp; Wilkins</general><scope>IQODW</scope><scope>CGR</scope><scope>CUY</scope><scope>CVF</scope><scope>ECM</scope><scope>EIF</scope><scope>NPM</scope><scope>AAYXX</scope><scope>CITATION</scope><scope>7TS</scope><scope>7X8</scope></search><sort><creationdate>20041201</creationdate><title>Limitations of ordinary least squares models in analyzing repeated measures data</title><author>UGRINOWITSCH, Carlos ; FELLINGHAM, Gilbert W ; RICARD, Mark D</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-c496t-13faad15032086dd9d84342e11a182486ac65be13be4611aff7ab3311589a34c3</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2004</creationdate><topic>Biological and medical sciences</topic><topic>Endpoint Determination</topic><topic>Exercise - physiology</topic><topic>Fundamental and applied biological sciences. Psychology</topic><topic>Humans</topic><topic>Least-Squares Analysis</topic><topic>Models, Theoretical</topic><topic>Reproducibility of Results</topic><topic>Space life sciences</topic><topic>Task Performance and Analysis</topic><topic>Treatment Outcome</topic><topic>Vertebrates: body movement. Posture. Locomotion. Flight. Swimming. Physical exercise. Rest. Sports</topic><topic>Weight Lifting</topic><toplevel>peer_reviewed</toplevel><toplevel>online_resources</toplevel><creatorcontrib>UGRINOWITSCH, Carlos</creatorcontrib><creatorcontrib>FELLINGHAM, Gilbert W</creatorcontrib><creatorcontrib>RICARD, Mark D</creatorcontrib><collection>Pascal-Francis</collection><collection>Medline</collection><collection>MEDLINE</collection><collection>MEDLINE (Ovid)</collection><collection>MEDLINE</collection><collection>MEDLINE</collection><collection>PubMed</collection><collection>CrossRef</collection><collection>Physical Education Index</collection><collection>MEDLINE - Academic</collection><jtitle>Medicine and science in sports and exercise</jtitle></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext</fulltext></delivery><addata><au>UGRINOWITSCH, Carlos</au><au>FELLINGHAM, Gilbert W</au><au>RICARD, Mark D</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Limitations of ordinary least squares models in analyzing repeated measures data</atitle><jtitle>Medicine and science in sports and exercise</jtitle><addtitle>Med Sci Sports Exerc</addtitle><date>2004-12-01</date><risdate>2004</risdate><volume>36</volume><issue>12</issue><spage>2144</spage><epage>2148</epage><pages>2144-2148</pages><issn>0195-9131</issn><eissn>1530-0315</eissn><coden>MSPEDA</coden><abstract>To a) introduce and present the advantages of linear mixed models using generalized least squares (GLS) when analyzing repeated measures data; and b) show how model misspecification and an inappropriate analysis using repeated measures ANOVA with ordinary least squares (OLS) methodology can negatively impact the probability of occurrence of Type I error. The effects of three strength-training groups were simulated. Strength gains had two slope conditions: null (no gain), and moderate (moderate gain). Ten subjects were hypothetically measured at five time points, and the correlation between measurements within a subject was modeled as compound symmetric (CS), autoregressive lag 1 (AR(1)), and random coefficients (RC). A thousand data sets were generated for each correlation structure. Then, each was analyzed four times--once using OLS, and three times using GLS, assuming the following variance/covariance structures: CS, AR(1), and RC. OLS produced substantially inflated probabilities of Type I errors when the variance/covariance structure of the data set was not CS. The RC model was less affected by the actual variance/covariance structure of the data set, and gave good estimates across all conditions. Using OLS to analyze repeated measures data is inappropriate when the covariance structure is not known to be CS. Random coefficients growth curve models may be useful when the variance/covariance structure of the data set is unknown.</abstract><cop>Hagerstown, MD</cop><pub>Lippincott Williams &amp; Wilkins</pub><pmid>15570152</pmid><doi>10.1249/01.mss.0000147580.40591.75</doi><tpages>5</tpages><oa>free_for_read</oa></addata></record>
fulltext fulltext
identifier ISSN: 0195-9131
ispartof Medicine and science in sports and exercise, 2004-12, Vol.36 (12), p.2144-2148
issn 0195-9131
1530-0315
language eng
recordid cdi_proquest_miscellaneous_67134290
source MEDLINE; Journals@Ovid LWW Legacy Archive; Journals@Ovid Complete
subjects Biological and medical sciences
Endpoint Determination
Exercise - physiology
Fundamental and applied biological sciences. Psychology
Humans
Least-Squares Analysis
Models, Theoretical
Reproducibility of Results
Space life sciences
Task Performance and Analysis
Treatment Outcome
Vertebrates: body movement. Posture. Locomotion. Flight. Swimming. Physical exercise. Rest. Sports
Weight Lifting
title Limitations of ordinary least squares models in analyzing repeated measures data
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2024-12-24T09%3A26%3A33IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-proquest_cross&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Limitations%20of%20ordinary%20least%20squares%20models%20in%20analyzing%20repeated%20measures%20data&rft.jtitle=Medicine%20and%20science%20in%20sports%20and%20exercise&rft.au=UGRINOWITSCH,%20Carlos&rft.date=2004-12-01&rft.volume=36&rft.issue=12&rft.spage=2144&rft.epage=2148&rft.pages=2144-2148&rft.issn=0195-9131&rft.eissn=1530-0315&rft.coden=MSPEDA&rft_id=info:doi/10.1249/01.mss.0000147580.40591.75&rft_dat=%3Cproquest_cross%3E17485675%3C/proquest_cross%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_pqid=17485675&rft_id=info:pmid/15570152&rfr_iscdi=true