Greetings Folks,
I'm currently running some iterations of some Hormone/Molecular data from my research on turtle hormone concentrations.
At the moment I've grouped my data into the three groupings based on Sampling Period.
There are three different Variables I am investigating, Estrogen, Testosterone, and Vitellogenin Concentration of plasma. Each of categories has varying sample size due to this being a field based project.
When I run ANOVA, none of the variables meet BOTH typical assumptions (Vitellogenin meets distribution but not the assumption of equal variance, Estrogen meets neither, etc). Thus I've concluded non-parametric analyses might be better to compare concentrations between these months.
I've run Kruskal-Wallace, and it suggests significance each time for two variables (Estrogen and Vitellogenin), but when I look at Cover-Inmann analysis it shows no significance being present, then when looking at Dwass-Steel-Chritchlow-Fligner for the same values, the P-values all appear significant.
Perhaps I am misinterpreting the results of these? Any advice on where to go next would be greatly appreciated!
Sincerely,
Jordan Donini
You could also try Mann-Whitney U-Tests. You need to use some of these non=parametric methods and then present the data as Medians and Quartiles, Rather than Means and SE.
Another possibility is to do non-parametric bootsraps of the mean for each group to generate 95 % confidence intervals (95% CIs) and then determine if the 95 % CIs overlap for the different groups. I've got some R code for this, but first you might Google a little if you think this might be something you would like to try.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA