Dear All,
My lab primarily works on insect wing patterns. In one of the projects, my student and I have defined 19 abnormality characters on the forewing and 6 abnormality characters on the hindwing of honey bees and scored them on both right and left wings in 14 populations with a total sample size of 424. Vein fusion, duplication, loss, etc. are considered abnormalities on insect wings. Now we have a table with 2 columns (right-left) and 19 rows (characters) for the forewing and a table with 2 columns (right-left) and 6 rows (characters) for the hindwing and the data is the count of individuals showing that particular abnormality. The question is whether these abnormalities occur symmetrical or nonsymmetrical on forewings and hindwings.
A simple graph of the data shows symmetrical occurrence but We thought we might back it up with some statistics to avoid arbitrary conclusions. Since we have 2 columns of Data, we thought we could use t-test to compare right and left forewing, and right and left hindwing separately. We ran a normality test as a precondition for t-test but the results showed that the data is not normally distributed and we cannot use t-test. So we went for the nonparametric Mann-Whitney U test instead and saw that, as expected based on the graph, there are no significant differences between the right and left forewing and right and left hindwing in those scored characters and abnormalities seem to happen symmetrical. We used PAST software package for the analyses. Now one of the reviewers of the manuscript insisted on Bonferroni corrected p-values for the analysis.
My first question is whether it is essential to use Bonferroni corrected p-values?
My second question is if there is a software package that we could use to get Bonferroni corrected p-values for Mann-Whitney U test?
I appreciate your answers in advance
You want to demonstrate that the abnormalities occur symmetrically, and you show this in a graph. That's ok and it should be sufficient.
The rest you describe is weird and wrong. Neither t-test nor U-test are able to show that the occurrences are symmetric. They answer completely different questions that are not interesting in this context.
You may, if you insist to "test" something, find some statistic that actually measures "symmetry" in some way and apply a test on that. The null hypothesis should not be "perfect symmetry" but rather the minimum relevant asymmetry. You may equivalently check if the confidence interval of your "symmetry"-statistic is fully withing the "range of negligibility". Such a confidence interval could be obtained by bootstrapping.
The reviewer has not understood anything and is just writing rubbish.
Are the values you are comparing counts of abnormalities ?
I don't quite understand your set-up, but perhaps something like a binomial test (similar to a chi-square goodness-of-fit test with two categories) for each abnormality... (?)
I'm not sure what the reviewer thinks that a Bonferroni correction would accomplish. If you're already seeing non-significant p-values, adding a correction won't make them significant...
Sal Mangiafico Thank you very much for your answer, Yes, the values are counts. Vein fusion, duplication, loss, etc. are considered abnormalities on insect wings.
Let's say abnormality 1 is an extra vein in a special position on the forewing, we counted how many bees out of 424 bees show that particular abnormality on their right forewing, and how many bees show that abnormality on their left forewing.
Jochen Wilhelm Thank you very much for your answer, We wanted to back up our findings with some statistics to avoid arbitrary conclusions. The values for scored characters are close to each other (not exact match) on the right and left wings. I appreciate any particular test suggestion.
The particular test suggestion is to identify a statistic that meaningfully measures "symmetry" and then to bootstrap its confidence interval. The "test" is to show that the interval is fully within a range considered "essentially symmetrical" (or as "range of no or negligible asymmetry").
- Badges
- Science topic
- Similar topics
- Filing