I am treating populations A and B with drug X and with drug Y. I get a significant difference in response to drug X between populations A and B via the Mann Whitney U test. Meanwhile, there is no significant difference in response to drug Y between A and B via the Mann-Whitney U test However it shows a significant difference via the student-t-test.
Can I report that both drugs X and Y are giving significant differences in response between populations A and B via Mann-Whitney and Student T-test?
Any input is highly appreciable .
Thanks
Hello Suresh,
The first thing to decide is, which of the tests is better suited for your research question and the characteristics of the scores that you're comparing?
Note that the parametric t-test and the non-parametric Mann-Whitney test are evaluating different null hypotheses about your data (t-test: equality of population means, given otherwise identical population distributions; Mann-Whitney: equality of distributions...if you can assume symmetric distributions, the M-W becomes a test of equality of medians). Both require homogeneity of variance and independence of observations within and across groups.
So, I'd select the one that seems plausible (given the points listed above) and report it for both comparisons.
Finally, I have seen people report the t-test for data that might not have fully met the usual assumptions (e.g., normality), then also reported the M-W as corroboration of the result (of course, corroboration implies that the two tests yielded the same conclusion about the respective null hypotheses).
Good luck with your work.
Many thanks David Morse.
Consider this situation, suppose my research question is
1. Do Heights (X) show any difference between two different populations A (Black) and B (White)
2. Do Weights (Y ) show any difference between two different populations A (Black) and B (White)
In my case distribution of X (Heights )and Y (Heights) is not uniform, it's skewed for both populations A and B. So Ideally I should do the Mann-Whitney U test for X and Y.
However, I applied the central limit theorem to Y (Heights) and choose the Independent t test, as my population is >35 in each group (A and B).
Whereas for X (Weights) I did not apply the central limit theorem, though my population is >35 in each group A and B . However, I choose the Mann-Whitney U test and not the Independent t-test, bcoz Independent-t-test is not giving significance (LOL). Only the Mann-Whitney U test is giving significance. (LOL)
My query is
Can I report in my manuscript that X shows a sig difference based on the Independent t-test, but not for the Maan-Whitney U test.
Similarly, Y shows a sig difference based on Mann-Whitney U test but not via the independent -t test.
Finally, there is some sig for both parameters X (Height) and Y (Weight) in between populations A (Balck) and B (White).
Does it make sense? applying the central limit theorem for Y (Weights) and not for X (Heights).
THANKS
- Badges
- Science topic
- Similar topics
- Chemistry
- Pharmacology
- Pharmaceuticals
- Drugs