Hi Wolfgang, The best methods for binary rare events is controversial. Some have argued for RD, but most opinion favors relative measures (RR, OR). Yilei Yu and Lu Tian published the exactmeta package a few years ago for handling rare events with an exact fixed effect model (either RD or RR). Will the metafor function permutest allows a similar analysis? Nathan
[R-meta] exact binomial meta analysis
3 messages · Nathan Leon Pace, MD, MStat, Viechtbauer Wolfgang (STAT), Ken Beath
I think conceptually permutest() does something different than the method in exactmeta, but I would have to study Tian et al. (2009) more carefully to really understand what they are doing. There are also 'exact' methods based on permutations of the 2x2 table data themselves (as opposed to permutest(), which in essence permutes the row labels of the tables). I am thinking of the work by Mehta and colleagues, like: Mehta, C. R., Patel, N. R., & Gray, R. (1985). Computing an exact confidence interval for the common odds ratio in several 2x2 contingency tables. Journal of the American Statistical Association, 80(392), 969-973. I think this kind of stuff can be found in Statexact (http://www.cytel.com/software/statxact). Implementing this in R would be quite interesting, but also challenging. Best, Wolfgang -----Original Message----- From: Nathan Pace [mailto:n.l.pace at utah.edu] Sent: Monday, August 28, 2017 23:40 To: Viechtbauer Wolfgang (SP); r-sig-meta-analysis at r-project.org Subject: exact binomial meta analysis Hi Wolfgang, The best methods for binary rare events is controversial. Some have argued for RD, but most opinion favors relative measures (RR, OR). Yilei Yu and Lu Tian published the exactmeta package a few years ago for handling rare events with an exact fixed effect model (either RD or RR). Will the metafor function permutest allows a similar analysis? Nathan
A Bayesian approach seems like a good option, provided that the odds ratio is fine. It could be fairly quickly implemented in rstan as it is just the random effects code without the random effect. It is argued that a fixed effects model may be optimistic. Another alternative is a generalized linear model using profile likelihood for the confidence intervals. Ken
On 30 Aug 2017, at 12:35 am, Viechtbauer Wolfgang (SP) <wolfgang.viechtbauer at maastrichtuniversity.nl> wrote: I think conceptually permutest() does something different than the method in exactmeta, but I would have to study Tian et al. (2009) more carefully to really understand what they are doing. There are also 'exact' methods based on permutations of the 2x2 table data themselves (as opposed to permutest(), which in essence permutes the row labels of the tables). I am thinking of the work by Mehta and colleagues, like: Mehta, C. R., Patel, N. R., & Gray, R. (1985). Computing an exact confidence interval for the common odds ratio in several 2x2 contingency tables. Journal of the American Statistical Association, 80(392), 969-973. I think this kind of stuff can be found in Statexact (http://www.cytel.com/software/statxact). Implementing this in R would be quite interesting, but also challenging. Best, Wolfgang -----Original Message----- From: Nathan Pace [mailto:n.l.pace at utah.edu] Sent: Monday, August 28, 2017 23:40 To: Viechtbauer Wolfgang (SP); r-sig-meta-analysis at r-project.org Subject: exact binomial meta analysis Hi Wolfgang, The best methods for binary rare events is controversial. Some have argued for RD, but most opinion favors relative measures (RR, OR). Yilei Yu and Lu Tian published the exactmeta package a few years ago for handling rare events with an exact fixed effect model (either RD or RR). Will the metafor function permutest allows a similar analysis? Nathan
_______________________________________________ R-sig-meta-analysis mailing list R-sig-meta-analysis at r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-meta-analysis