If sample size is more than 100 and the data is non-parametrically distributed, which test is suitable- Independent t-test or Mann-Whitney test? Note that I’ve already run an analysis on both, and found a substantial difference between mean values, but not in p-values.
It depends on what kind of hypothesis you want to test. If you want to test the mean difference, then use the t-test; if you want to test stochastic equivalence, then use the U-test.
There are 4 assumptions for carrying out independent t test:
1. Random sampling
2. Independent observations
[1 and 2 are usually covered by the research method]
3. In each group, data are normally distributed (seen from histogram of each of the two groups) or total sample size is >=30
4. Population variance are the same between the two groups (seen from Levene’s Test for Equality of Variances—P value of >0.05 is acceptable—in the SPSS output).
so Mr Mamunur, if either of the 2 groups is >=30 sample size (or if either is 0.05 (from SPSS output), independent t test is appropriate.
Hope this helps!
Can you clairify what your mean by ' already run an analysis on both, and found a substantial difference between mean values, but not in p-values '?
a) you have ran both T and U tests and they give different group values to each other but the p value was almost identical between the tests
b) you have ran both T and U tests and not got a significant p value with either but the difference between mean values for your groups is large so you expected at least one of the tests to be significant.
If your data is following non-normal distribution, then you must go for Mann whitney U test instead of independent t test.
It depends on what kind of hypothesis you want to test. If you want to test the mean difference, then use the t-test; if you want to test stochastic equivalence, then use the U-test.
Dear you have been already done both method and not found any diffrence in p value but ethically and to get the accurate results in such typeof condition we should used only those method go with assumption and as per condition it is clearly predict that is most suitable test than t test. Mann-Whitney test
Hi James! I mean t-test and Mann-Whitney test were performed on same variables separately and found differences in mean values, but not in significant p-value meaning that what variables were significant in t test, that also remained in Mann-Whitney test. Hope you understand now!
Thank you all for your explanation!
Article Bayesian alternatives for common null-hypothesis significanc...
Article Equivalence Tests: A Practical Primer for t Tests, Correlati...
https://www.ncbi.nlm.nih.gov/pmc/articles/pmc7275319/
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size