Application of generalized estimating equations and linear mixed effects models to analysis of correlated data in the field of publication bias in the reporting of randomized clinical trials
The research focused on identifying trial characteristics leading to delayed publication of randomized comparisons, and hence publication bias. Time to first mention in an article (irrespective of whether results are given) and to first reporting of results were modelled using ordinary linear regres...
| Main Authors: | , |
|---|---|
| Format: | Journal article |
| Language: | English |
| Published: |
2015
|
| Subjects: |
| _version_ | 1797067099928002560 |
|---|---|
| author | Burrett, J Lunn, D |
| author_facet | Burrett, J Lunn, D |
| author_sort | Burrett, J |
| collection | OXFORD |
| description | The research focused on identifying trial characteristics leading to delayed publication of randomized comparisons, and hence publication bias. Time to first mention in an article (irrespective of whether results are given) and to first reporting of results were modelled using ordinary linear regression (independence model). These analyses were extended to include all mentions and all reportings of results where non-independence necessitated using repeated measures techniques. The residuals from the independence model were used to construct a covariance matrix, thereby suggesting plausible correlation structures for repeated measures models. Results from two methods; generalized estimating equations (GEE) and linear mixed effects modelling, are compared. Problems caused by missing data and their solution are also discussed. This paper concentrates on methodology and the use of repeated measures techniques for incorporating appropriate correlation structures, rather than interpretation of findings, which is published separately. Application of the methods is described, as is the importance of the correct use of repeated measures analyses when an independence model is inappropriate; an independence model may approximate well to the final model, but should only be used to suggest useful correlation structures. Repeated measures methods are easily implemented, providing practical ways of dealing with correlated data. |
| first_indexed | 2024-03-06T21:51:28Z |
| format | Journal article |
| id | oxford-uuid:4b6bd83b-49db-4765-9a65-74f9d18dbc6c |
| institution | University of Oxford |
| language | English |
| last_indexed | 2024-03-06T21:51:28Z |
| publishDate | 2015 |
| record_format | dspace |
| spelling | oxford-uuid:4b6bd83b-49db-4765-9a65-74f9d18dbc6c2022-03-26T15:43:30ZApplication of generalized estimating equations and linear mixed effects models to analysis of correlated data in the field of publication bias in the reporting of randomized clinical trials Journal articlehttp://purl.org/coar/resource_type/c_dcae04bcuuid:4b6bd83b-49db-4765-9a65-74f9d18dbc6cArtStatisticsMedical SciencesEnglishOxford University Research Archive - Valet2015Burrett, JLunn, DThe research focused on identifying trial characteristics leading to delayed publication of randomized comparisons, and hence publication bias. Time to first mention in an article (irrespective of whether results are given) and to first reporting of results were modelled using ordinary linear regression (independence model). These analyses were extended to include all mentions and all reportings of results where non-independence necessitated using repeated measures techniques. The residuals from the independence model were used to construct a covariance matrix, thereby suggesting plausible correlation structures for repeated measures models. Results from two methods; generalized estimating equations (GEE) and linear mixed effects modelling, are compared. Problems caused by missing data and their solution are also discussed. This paper concentrates on methodology and the use of repeated measures techniques for incorporating appropriate correlation structures, rather than interpretation of findings, which is published separately. Application of the methods is described, as is the importance of the correct use of repeated measures analyses when an independence model is inappropriate; an independence model may approximate well to the final model, but should only be used to suggest useful correlation structures. Repeated measures methods are easily implemented, providing practical ways of dealing with correlated data. |
| spellingShingle | Art Statistics Medical Sciences Burrett, J Lunn, D Application of generalized estimating equations and linear mixed effects models to analysis of correlated data in the field of publication bias in the reporting of randomized clinical trials |
| title | Application of generalized estimating equations and linear mixed effects models to analysis of correlated data in the field of publication bias in the reporting of randomized clinical trials |
| title_full | Application of generalized estimating equations and linear mixed effects models to analysis of correlated data in the field of publication bias in the reporting of randomized clinical trials |
| title_fullStr | Application of generalized estimating equations and linear mixed effects models to analysis of correlated data in the field of publication bias in the reporting of randomized clinical trials |
| title_full_unstemmed | Application of generalized estimating equations and linear mixed effects models to analysis of correlated data in the field of publication bias in the reporting of randomized clinical trials |
| title_short | Application of generalized estimating equations and linear mixed effects models to analysis of correlated data in the field of publication bias in the reporting of randomized clinical trials |
| title_sort | application of generalized estimating equations and linear mixed effects models to analysis of correlated data in the field of publication bias in the reporting of randomized clinical trials |
| topic | Art Statistics Medical Sciences |
| work_keys_str_mv | AT burrettj applicationofgeneralizedestimatingequationsandlinearmixedeffectsmodelstoanalysisofcorrelateddatainthefieldofpublicationbiasinthereportingofrandomizedclinicaltrials AT lunnd applicationofgeneralizedestimatingequationsandlinearmixedeffectsmodelstoanalysisofcorrelateddatainthefieldofpublicationbiasinthereportingofrandomizedclinicaltrials |