Which type of statistical analysis to run in R with 2 treatment types in my study?

Question

(Originally asked on Stack Overflow, reposting here after a commentor mentioned this is a better place to ask it.)

I have a seed germination study that I am trying to analyze. In this study I had 4 different media types (nutritious agar plus different vitamins/additives) with 40 plates each. Each group of 40 plates was split into 5 groups of 8 plates to further test sterilization/scarification pretreatment times (10, 30, 60, 90, 120 minutes). The resulting germination of all seed per plate was counted into 6 different stages of growth (stages 0-5, stage 0 representing no germination of the seed).I used Excel to show me the percent germination per plate (stages 1-5). Technically stages 3-5 are more important, but I am just looking at the ability to germinate so for the sake of this study stages 1-5 are fine to group together to give the binary of yes or no on germination. My question is what statistical test should I be running in R? I've included a screenshot of my data.

I've played around with Python before deciding to use R. At the end of the day, I've realized that I just don't know which test I should be running: ANOVA, 2-Way ANOVA, MANOVA, etc... My stats skills have a ways to go, this is me trying to learn.

Answer 1

If your dependent variable (DV) is binary (germinated vs. didn't) then you don't want any kind of ANOVA. That's for continuous DV.

You have multiple time points. From the data, it looks like different seeds at different time points. If that's the case, then you can do logistic regression with media and time as independent variables.

If they are the same seeds tested for germination at multiple time points, then I'd say a nonlinear multilevel model could be good.

You say that it's fine to lump the stages -- OK. But then, why did you collect it? If you want to not lump the stages, you can use ordinal logistic.