0

Mice were exposed to light for 0, 1, and 4 hours. I have an algorithm that calculates their cell embeddings depending on the duration of exposure. As a result, violin plots of my data look like this:

enter image description here

my independent variable (cell embeddings) has negative and positive values. What would be the best way to find if stimulation significantly affected mice's gene expression? I assume I can't use ANOVA cause my data is not normally distributed.

Yulia Kentieva
  • 119
  • 2
    Your data is significantly skewed so you really can't use ANOVA. In contrast, Kruskal–Wallis one-way analysis of variance doesn't not need any distributional assumptions. However, "Kruskal-Wallis is more general than a comparison of means: it tests whether the probability that a random observation from each group is equally likely to be above or below a random observation from another group." look at this answer: https://stats.stackexchange.com/a/76080/155375 – The_old_man Jan 11 '23 at 16:48
  • 2
    You'll need a clear, precise definition of "affected"... for example, "affected" could mean the mean cell embedding changed, or the probability of a positive cell embedding changed, or cell embeddings were "probably" larger in general (as with the Kruskal-Wallis test @The_old_man mentioned(+1)). That will help determine what test to use. – jbowman Jan 11 '23 at 17:43

0 Answers0