Re: [R] Is there an exact binomial k-sample test of equivalence?
2011/3/11 Albyn Jones jo...@reed.edu but presumably what you really want would be based on a joint confidence region for all the proportions. I've had read On Exact Methods for Testing Equality of Binomial Proportions by Akihito Matsuo, but still, this concept is for me unclear and I got lost... We have H0: |pi1-pi2-pi3| = 0.05 n1-40;n2-40;n3-40 s1-list(1:11);s2-list(1:17);s3-list(1:15) pi1-max(s1[[1]])/n1;pi2-max(s2[[1]])/n2;pi3-max(s3[[1]])/n3 epsilon-.05 t(c(pi1,pi2,pi3)) T_chi.sq-sum(sapply(s1,(function(s1){(s1-n1*pi1)^2/n1*pi1*(1-pi1)}))) T_binom-sum(sapply(s1,function(s1){choose(n1,s1)*(pi1-epsilon)^s1*(pi1+epsilon)^(n1-s1)})) # Or it's about sum of all s-list(1:43) and n-n1+n2+n3 ? Am I going to right direction? -- Mi³ego dnia [[alternative HTML version deleted]] __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
[R] Is there an exact binomial k-sample test of equivalence?
Hi, I've got one silly question for evening. I don't know is this reasonable, but can test with two the most extreme proportions from the samples could be good enough evidence for testing equivalence, or should I have to look for something else...? -- Mi³ego dnia [[alternative HTML version deleted]] __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] Is there an exact binomial k-sample test of equivalence?
Presumably the null hypothesis is that at least one of the differences is larger in absolute magnitude than the chosen epsilon. I expect that your procedure would be conservative: if it rejects the null hypothesis, then you are ok, but presumably what you really want would be based on a joint confidence region for all the proportions. albyn Quoting ?ukasz R?c?awowicz lukasz.reclawow...@gmail.com: Hi, I've got one silly question for evening. I don't know is this reasonable, but can test with two the most extreme proportions from the samples could be good enough evidence for testing equivalence, or should I have to look for something else...? -- Mi³ego dnia [[alternative HTML version deleted]] __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.