Sample size determination for logistic regression revisited.
E. Demidenko. Statistics in medicine, 26 (18):
3385-97(August 2007)4918<m:linebreak></m:linebreak>LR: 20071115; JID: 8215016; ppublish;<m:linebreak></m:linebreak>Sample size; Regressió logística.
DOI: 10.1002/sim.2771
Abstract
There is no consensus on the approach to compute the power and sample size with logistic regression. Some authors use the likelihood ratio test; some use the test on proportions; some suggest various approximations to handle the multivariate case. We advocate the use of the Wald test since the Z-score is routinely used for statistical significance testing of regression coefficients. The null-variance formula became popular from early studies, which contradicts modern software, which utilizes the method of maximum likelihood estimation (MLE), when the variance of the MLE is estimated at the MLE, not at the null. We derive general Wald-based power and sample size formulas for logistic regression and then apply them to binary exposure and confounder to obtain a closed-form expression. These formulas are applied to minimize the total sample size in a case-control study to achieve a given power by optimizing the ratio of controls to cases. Approximately, the optimal number of controls to cases is equal to the square root of the alternative odds ratio. Our sample size and power calculations can be carried out online at www.dartmouth.edu/ approximately eugened.
%0 Journal Article
%1 Demidenko2007
%A Demidenko, Eugene
%D 2007
%J Statistics in medicine
%K ClinicalTrialsasTopic ClinicalTrialsasTopic:statistics&numericald LogisticModels SampleSize UnitedStates
%N 18
%P 3385-97
%R 10.1002/sim.2771
%T Sample size determination for logistic regression revisited.
%U http://www.ncbi.nlm.nih.gov/pubmed/17149799
%V 26
%X There is no consensus on the approach to compute the power and sample size with logistic regression. Some authors use the likelihood ratio test; some use the test on proportions; some suggest various approximations to handle the multivariate case. We advocate the use of the Wald test since the Z-score is routinely used for statistical significance testing of regression coefficients. The null-variance formula became popular from early studies, which contradicts modern software, which utilizes the method of maximum likelihood estimation (MLE), when the variance of the MLE is estimated at the MLE, not at the null. We derive general Wald-based power and sample size formulas for logistic regression and then apply them to binary exposure and confounder to obtain a closed-form expression. These formulas are applied to minimize the total sample size in a case-control study to achieve a given power by optimizing the ratio of controls to cases. Approximately, the optimal number of controls to cases is equal to the square root of the alternative odds ratio. Our sample size and power calculations can be carried out online at www.dartmouth.edu/ approximately eugened.
%@ 0277-6715
@article{Demidenko2007,
abstract = {There is no consensus on the approach to compute the power and sample size with logistic regression. Some authors use the likelihood ratio test; some use the test on proportions; some suggest various approximations to handle the multivariate case. We advocate the use of the Wald test since the Z-score is routinely used for statistical significance testing of regression coefficients. The null-variance formula became popular from early studies, which contradicts modern software, which utilizes the method of maximum likelihood estimation (MLE), when the variance of the MLE is estimated at the MLE, not at the null. We derive general Wald-based power and sample size formulas for logistic regression and then apply them to binary exposure and confounder to obtain a closed-form expression. These formulas are applied to minimize the total sample size in a case-control study to achieve a given power by optimizing the ratio of controls to cases. Approximately, the optimal number of controls to cases is equal to the square root of the alternative odds ratio. Our sample size and power calculations can be carried out online at www.dartmouth.edu/ approximately eugened.},
added-at = {2023-02-03T11:44:35.000+0100},
author = {Demidenko, Eugene},
biburl = {https://www.bibsonomy.org/bibtex/24690c893f48c2d1b81974992e25b2539/jepcastel},
city = {Dartmouth Medical School, Hanover, NH 03755, USA. eugened@dartmouth.edu},
doi = {10.1002/sim.2771},
interhash = {f0731f674217eb96b0b4bdb02e5fe2b2},
intrahash = {4690c893f48c2d1b81974992e25b2539},
isbn = {0277-6715},
issn = {0277-6715},
journal = {Statistics in medicine},
keywords = {ClinicalTrialsasTopic ClinicalTrialsasTopic:statistics&numericald LogisticModels SampleSize UnitedStates},
month = {8},
note = {4918<m:linebreak></m:linebreak>LR: 20071115; JID: 8215016; ppublish;<m:linebreak></m:linebreak>Sample size; Regressió logística},
number = 18,
pages = {3385-97},
pmid = {17149799},
timestamp = {2023-02-03T11:44:35.000+0100},
title = {Sample size determination for logistic regression revisited.},
url = {http://www.ncbi.nlm.nih.gov/pubmed/17149799},
volume = 26,
year = 2007
}