I am trying to get two sample sizes (treatment vs. control) for differences in proportions (e.g. 80% vs. 70% of success), for non-inferiority and equivalence trials with R-package TrialSize. Actually I have two questions:
1) when writing the formula, both for equivalence and non-inferiority design, I am asked to insert "superiority or non-inferiority margin": I know this should be set so the alternative hypothesis is |p1 - p2| > margin, but I don't know how should it be set. Is there a rule?
2) if I set "superiority or non-inferiority margin" at 0.005, for example, and compute the sample size for equivalence trial and non-inferiority trial, I get the same sample size!!! So should this "margin" be kept lower in non-inferiority design than in equivalence design?
I know there should be something wrong, but I cannot find where the mistake is!
Thanks to everybody for answering :-)
Dear Sir,
I propose PASS software to determine sample size in each task:
http://www.ncss.com/download/pass/free-trial/
best regards
The answer to question 1 is clinical judgement. Just as in determining the clinically important effect size, subject-matter expertise is required to judge what constitutes clinical agreement.
You might also look for the G*Power 3 software at http://www.gpower.hhu.de/
There's no reason why you can't use TrialSize, though. Each test has alternative versions designed for equivalence and non inferiority trials, for example:
One Sample Mean Test for Equivalence
Ho: |margin| ≥ delta Ha: |margin| < delta
The test is concluded to be equivalent to a gold standard on average if the null hypothesis is rejected at significance level alpha
I think you will find the package easier to use if you borrow the accompanying text from the library (Chapman and Hall, Sample Size Calculations in Clinical Research, Chow SC, Shao J, Wang H.). But I should warn you that there are typos in some of the derivations, which make it hard to follow. For this reason, I prefer Sample Sizes for Clinical Trials by Steven A. Julious.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Applied Statistics