Hello everyone,
I ran in some issues, when i wanted to test my data. My data consists of 7 monkeys with the frequencies of stress behaviour, self grooming (SG) and stereotypical behaviour (ST).
I checked the assumptions for the ANOVA test and non were met (see attached pictures). Therefore, i wanted to perform an alternative test: the Kruskall Wallis test. I read that all groups should have the same shape distributions. This is again not the case in my data.
Now i dont know which test i could use (and how to execute it in R) and i was hoping someone could help me with my problem.
thanks in advance!
What are your hypotheses? Are you interested in comparing those outcomes between each monkey?
Joe Stanek yes, i want to compare the outcomes between each monkey
I assume you have several data points per monkey and variable (SG and ST). Say you want to knoo if there are significant differences between monkeys.
The non parametric apprach is Kruskal Wallis. In R you code this:
x
Víctor Morales-Oñate Thanks you for the help! I got the results.
However, now I realised that I also want to know if the SG and ST (per monkey) are differing from 0. I couldn't find out if the Dunn test could give me the 95% confidence interval. Is it possible to get the 95& CI or should I do another test?
In that case, keeping in the non parametric world, you can use a one sample Wilcoxon test. In R, you need to do the following per monkey and variable:
wilcox.test(SG, mu = 0)
That means that for SG i have to execute 7 tests, because i have 7 monkeys? Then i should also use a correction, because of the multiple testing?
For multiple testing I suggest the Bonferroni correction because it is simple.
Details in https://en.wikipedia.org/wiki/Bonferroni_correction
These are count data so I'd think about model this with a Poisson type model (actually probably quasipoisson as they look over dispersed and possibly zero inflated).
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA