Comparison of Logistic Regression versus Propensity Score When the Number of Events Is Low and There Are Multiple Confounders. Cepeda, M. S., Boston, R., Farrar, J. T., & Strom, B. L. Am J Epi, 158:280-287, 2003. doi abstract bibtex The aim of this study was to use Monte Carlo simulations to compare logistic regression with propensity scores in terms of bias, precision, empirical coverage probability, empirical power, and robustness when the number of events is low relative to the number of confounders. The authors simulated a cohort study and performed 252,480 trials. In the logistic regression, the bias decreased as the number of events per confounder increased. In the propensity score, the bias decreased as the strength of the association of the exposure with the outcome increased. Propensity scores produced estimates that were less biased, more robust, and more precise than the logistic regression estimates when there were seven or fewer events per confounder. The logistic regression empirical coverage probability increased as the number of events per confounder increased. The propensity score empirical coverage probability decreased after eight or more events per confounder. Overall, the propensity score exhibited more empirical power than logistic regression. Propensity scores are a good alternative to control for imbalances when there are seven or fewer events per confounder; however, empirical power could range from 35% to 60%. Logistic regression is the technique of choice when there are at least eight events per confounder.
@article{cep03com,
title = {Comparison of Logistic Regression versus Propensity Score When the Number of Events Is Low and There Are Multiple Confounders},
volume = {158},
abstract = {The aim of this study was to use Monte Carlo simulations to compare logistic regression with propensity scores in terms of bias, precision, empirical coverage probability, empirical power, and robustness when the number of events is low relative to the number of confounders. The authors simulated a cohort study and performed 252,480 trials. In the logistic regression, the bias decreased as the number of events per confounder increased. In the propensity score, the bias decreased as the strength of the association of the exposure with the outcome increased. Propensity scores produced estimates that were less biased, more robust, and more precise than the logistic regression estimates when there were seven or fewer events per confounder. The logistic regression empirical coverage probability increased as the number of events per confounder increased. The propensity score empirical coverage probability decreased after eight or more events per confounder. Overall, the propensity score exhibited more empirical power than logistic regression. Propensity scores are a good alternative to control for imbalances when there are seven or fewer events per confounder; however, empirical power could range from 35\% to 60\%. Logistic regression is the technique of choice when there are at least eight events per confounder.},
journal = {Am J Epi},
doi = {10.1093/aje/kwg115},
author = {Cepeda, M. S. and Boston, R. and Farrar, J. T. and Strom, B. L.},
year = {2003},
pages = {280-287},
eprint = {http://aje.oxfordjournals.org/content/158/3/280.full.pdf+html},
citeulike-article-id = {13265471},
citeulike-linkout-0 = {http://dx.doi.org/10.1093/aje/kwg115},
citeulike-linkout-1 = {http://aje.oxfordjournals.org/content/158/3/280.abstract},
posted-at = {2014-07-14 14:09:57},
priority = {0},
annote = {propensity score analysis has advantages to traditional outcome modeling in this case}
}
Downloads: 0
{"_id":"35LwS3vRCH3q6Jygo","bibbaseid":"cepeda-boston-farrar-strom-comparisonoflogisticregressionversuspropensityscorewhenthenumberofeventsislowandtherearemultipleconfounders-2003","downloads":0,"creationDate":"2018-06-23T20:06:35.136Z","title":"Comparison of Logistic Regression versus Propensity Score When the Number of Events Is Low and There Are Multiple Confounders","author_short":["Cepeda, M. S.","Boston, R.","Farrar, J. T.","Strom, B. L."],"year":2003,"bibtype":"article","biburl":"http://hbiostat.org/bib/harrelfe.bib","bibdata":{"bibtype":"article","type":"article","title":"Comparison of Logistic Regression versus Propensity Score When the Number of Events Is Low and There Are Multiple Confounders","volume":"158","abstract":"The aim of this study was to use Monte Carlo simulations to compare logistic regression with propensity scores in terms of bias, precision, empirical coverage probability, empirical power, and robustness when the number of events is low relative to the number of confounders. The authors simulated a cohort study and performed 252,480 trials. In the logistic regression, the bias decreased as the number of events per confounder increased. In the propensity score, the bias decreased as the strength of the association of the exposure with the outcome increased. Propensity scores produced estimates that were less biased, more robust, and more precise than the logistic regression estimates when there were seven or fewer events per confounder. The logistic regression empirical coverage probability increased as the number of events per confounder increased. The propensity score empirical coverage probability decreased after eight or more events per confounder. Overall, the propensity score exhibited more empirical power than logistic regression. Propensity scores are a good alternative to control for imbalances when there are seven or fewer events per confounder; however, empirical power could range from 35% to 60%. Logistic regression is the technique of choice when there are at least eight events per confounder.","journal":"Am J Epi","doi":"10.1093/aje/kwg115","author":[{"propositions":[],"lastnames":["Cepeda"],"firstnames":["M.","S."],"suffixes":[]},{"propositions":[],"lastnames":["Boston"],"firstnames":["R."],"suffixes":[]},{"propositions":[],"lastnames":["Farrar"],"firstnames":["J.","T."],"suffixes":[]},{"propositions":[],"lastnames":["Strom"],"firstnames":["B.","L."],"suffixes":[]}],"year":"2003","pages":"280-287","eprint":"http://aje.oxfordjournals.org/content/158/3/280.full.pdf+html","citeulike-article-id":"13265471","citeulike-linkout-0":"http://dx.doi.org/10.1093/aje/kwg115","citeulike-linkout-1":"http://aje.oxfordjournals.org/content/158/3/280.abstract","posted-at":"2014-07-14 14:09:57","priority":"0","annote":"propensity score analysis has advantages to traditional outcome modeling in this case","bibtex":"@article{cep03com,\n title = {Comparison of Logistic Regression versus Propensity Score When the Number of Events Is Low and There Are Multiple Confounders},\n volume = {158},\n abstract = {The aim of this study was to use Monte Carlo simulations to compare logistic regression with propensity scores in terms of bias, precision, empirical coverage probability, empirical power, and robustness when the number of events is low relative to the number of confounders. The authors simulated a cohort study and performed 252,480 trials. In the logistic regression, the bias decreased as the number of events per confounder increased. In the propensity score, the bias decreased as the strength of the association of the exposure with the outcome increased. Propensity scores produced estimates that were less biased, more robust, and more precise than the logistic regression estimates when there were seven or fewer events per confounder. The logistic regression empirical coverage probability increased as the number of events per confounder increased. The propensity score empirical coverage probability decreased after eight or more events per confounder. Overall, the propensity score exhibited more empirical power than logistic regression. Propensity scores are a good alternative to control for imbalances when there are seven or fewer events per confounder; however, empirical power could range from 35\\% to 60\\%. Logistic regression is the technique of choice when there are at least eight events per confounder.},\n journal = {Am J Epi},\n doi = {10.1093/aje/kwg115},\n author = {Cepeda, M. S. and Boston, R. and Farrar, J. T. and Strom, B. L.},\n year = {2003},\n pages = {280-287},\n eprint = {http://aje.oxfordjournals.org/content/158/3/280.full.pdf+html},\n citeulike-article-id = {13265471},\n citeulike-linkout-0 = {http://dx.doi.org/10.1093/aje/kwg115},\n citeulike-linkout-1 = {http://aje.oxfordjournals.org/content/158/3/280.abstract},\n posted-at = {2014-07-14 14:09:57},\n priority = {0},\n annote = {propensity score analysis has advantages to traditional outcome modeling in this case}\n}\n\n","author_short":["Cepeda, M. S.","Boston, R.","Farrar, J. T.","Strom, B. L."],"key":"cep03com","id":"cep03com","bibbaseid":"cepeda-boston-farrar-strom-comparisonoflogisticregressionversuspropensityscorewhenthenumberofeventsislowandtherearemultipleconfounders-2003","role":"author","urls":{},"downloads":0},"search_terms":["comparison","logistic","regression","versus","propensity","score","number","events","low","multiple","confounders","cepeda","boston","farrar","strom"],"keywords":["*import"],"authorIDs":[],"dataSources":["mEQakjn8ggpMsnGJi"]}