Assessing Subgroup Effects with Binary Data: Can the Use of Different Effect Measures Lead to Different Conclusions?. White, I. R. & Elbourne, D. BMC Med Res Methodol, April, 2005. bibtex @article{whi05ass,
title = {Assessing Subgroup Effects with Binary Data: Can the Use of Different Effect Measures Lead to Different Conclusions?},
volume = {5},
number = {15},
journal = {BMC Med Res Methodol},
author = {White, Ian R. and Elbourne, Diana},
month = apr,
year = {2005},
citeulike-article-id = {13265806},
citeulike-linkout-0 = {http://www.biomedcentral.com/1471-2288/5/15},
posted-at = {2014-07-14 14:10:04},
priority = {0},
annote = {example of disagreement between variation in treatment effects over subgroups when using relative risks vs. odds ratio;large variation in baseline risk;"The test of interaction should therefore be applied to the effect measure which is least likely to exhibit an interaction.";doi:10.1186/1471-2288-5-15;note from Doug Altman:"I fully support the idea of using evidence as well as theoretical arguments in such situations. We have evidence that OR is not more likely to give homogeneity in meta-analysis than RR (Deeks JJ: Issues in the selection of a summary statistic for meta-analysis of clinical trials with binary outcomes. Stat Med 2002, 21:1575-1600.). It certainly is possible for RRs to be constant across a wide range of risk groups when one is heading towards zero so to speak - ie a beneficial effect gives RR{$<$}1. I agree that it is not possible when one goes the other way."}
}
Downloads: 0
{"_id":"joDNSpcbcvLJ36By9","bibbaseid":"white-elbourne-assessingsubgroupeffectswithbinarydatacantheuseofdifferenteffectmeasuresleadtodifferentconclusions-2005","downloads":0,"creationDate":"2018-06-23T20:06:34.396Z","title":"Assessing Subgroup Effects with Binary Data: Can the Use of Different Effect Measures Lead to Different Conclusions?","author_short":["White, I. R.","Elbourne, D."],"year":2005,"bibtype":"article","biburl":"http://hbiostat.org/bib/harrelfe.bib","bibdata":{"bibtype":"article","type":"article","title":"Assessing Subgroup Effects with Binary Data: Can the Use of Different Effect Measures Lead to Different Conclusions?","volume":"5","number":"15","journal":"BMC Med Res Methodol","author":[{"propositions":[],"lastnames":["White"],"firstnames":["Ian","R."],"suffixes":[]},{"propositions":[],"lastnames":["Elbourne"],"firstnames":["Diana"],"suffixes":[]}],"month":"April","year":"2005","citeulike-article-id":"13265806","citeulike-linkout-0":"http://www.biomedcentral.com/1471-2288/5/15","posted-at":"2014-07-14 14:10:04","priority":"0","annote":"example of disagreement between variation in treatment effects over subgroups when using relative risks vs. odds ratio;large variation in baseline risk;\"The test of interaction should therefore be applied to the effect measure which is least likely to exhibit an interaction.\";doi:10.1186/1471-2288-5-15;note from Doug Altman:\"I fully support the idea of using evidence as well as theoretical arguments in such situations. We have evidence that OR is not more likely to give homogeneity in meta-analysis than RR (Deeks JJ: Issues in the selection of a summary statistic for meta-analysis of clinical trials with binary outcomes. Stat Med 2002, 21:1575-1600.). It certainly is possible for RRs to be constant across a wide range of risk groups when one is heading towards zero so to speak - ie a beneficial effect gives RR$<$1. I agree that it is not possible when one goes the other way.\"","bibtex":"@article{whi05ass,\n title = {Assessing Subgroup Effects with Binary Data: Can the Use of Different Effect Measures Lead to Different Conclusions?},\n volume = {5},\n number = {15},\n journal = {BMC Med Res Methodol},\n author = {White, Ian R. and Elbourne, Diana},\n month = apr,\n year = {2005},\n citeulike-article-id = {13265806},\n citeulike-linkout-0 = {http://www.biomedcentral.com/1471-2288/5/15},\n posted-at = {2014-07-14 14:10:04},\n priority = {0},\n annote = {example of disagreement between variation in treatment effects over subgroups when using relative risks vs. odds ratio;large variation in baseline risk;\"The test of interaction should therefore be applied to the effect measure which is least likely to exhibit an interaction.\";doi:10.1186/1471-2288-5-15;note from Doug Altman:\"I fully support the idea of using evidence as well as theoretical arguments in such situations. We have evidence that OR is not more likely to give homogeneity in meta-analysis than RR (Deeks JJ: Issues in the selection of a summary statistic for meta-analysis of clinical trials with binary outcomes. Stat Med 2002, 21:1575-1600.). It certainly is possible for RRs to be constant across a wide range of risk groups when one is heading towards zero so to speak - ie a beneficial effect gives RR{$<$}1. I agree that it is not possible when one goes the other way.\"}\n}\n\n","author_short":["White, I. R.","Elbourne, D."],"key":"whi05ass","id":"whi05ass","bibbaseid":"white-elbourne-assessingsubgroupeffectswithbinarydatacantheuseofdifferenteffectmeasuresleadtodifferentconclusions-2005","role":"author","urls":{},"downloads":0},"search_terms":["assessing","subgroup","effects","binary","data","use","different","effect","measures","lead","different","conclusions","white","elbourne"],"keywords":["*import"],"authorIDs":[],"dataSources":["mEQakjn8ggpMsnGJi"]}