Hey all, Iím new here.

I have an issue with a percentages and the use of ANOVA. But letís start from the beginning. I m a PhD Student and I study primary human granulocytes isolated from volunteers. The isolated granulocytes are treated with different compounds or cytokines (in total I have 6 groups). My readout is survival or viability of cells, which I assess by flow cytometry using Annexin V & PI double staining. Annexin V or PI single positives or double positives are apoptotic, while cells, which are negative for both are viable. Basically, I have 4 quadrants, which add up to 100%, obviously, and the one Iím interested in, the one with double negatives cells, is the percental survival of the cells. However, this is not a classic survival analysis, its not a binary model, were there is death or life, its just that the phenotype of the cells is indicative of its status. And by the way I acquire around 5000-8000 cells per group (grand total of all quadrants).
This looks something like this: https://www.google.ch/search?q=survi...RO9uFHbZgeM%3A

This kind of experiments is repeated with multiple donors. In the end, I have analysed the 6 six treatment groups with cells from different individuals. Up to now, I just presented my data as scatter plot with SD and I compared the different groups using an ordinary one-way ANOVA (using Prism), these were truly happy days (and thatís how most of the people using this assay present it, but that is not a good reason, I know). However, I recently read that percentage are bounded variables and therefore are not normally distributed. Since my biological replicates can be counted on one hand (however, it is to note that every data point in percentage I collect is based on the analysis of thousands of cells), I cannot really test whether my data is really normally distributed or not. I wrote a lot, so I going to ask my first question.
How do I present data like this? Is it okay to present my data as mean plus SD? The problem I have with the SD is that I sometimes have a mean of 4% survival and a SD of 9, meaning that the SD reaches below zero, which doesnít make any sense, since negative values cannot be measured. I read that one could use poisson CI for percentage, however, I have no idea how this could be presented in Prism.
Second question: Is one-way ANOVA the right way to compare data like this? I tried binomial test in Prism. I got rid of percentage and just put the number of events in the double negative quadrant in an "Alive group" and added the events in the remaining 3 quadrants in a group "Dead". Okay so now it gets really confusing for me. I have 6 treatment groups, 2 states (dead or alive), which have discrete values (number of cells) and 6 biological replicates. Since the cells originate from different individuals, the total count may vary a lot; however, the ratio between alive and dead cells between the groups is pretty obvious. But I have no idea if ratio can be compared between multiple groups consisting of multiple biological replicates. And thatís also the problem I have when I try binomial test, the program wants to know the number of subjects with one of these two outcomes (numerator) and the total number of subjects (denumerator), but then where can I enter my biological replicates? Furthermore, it is often not possible to compare multiple groups with tests like these. I also tested non-parametric tests like Kruskal-Wallis test, however, the ranking doesnít seem right, since the difference intra-donor difference messes up the ranking between groups (e.g. in the case donor A is low in group 1 and group 2 and donor B is high in group 1 and group 2), therefore for this kind of data, analysis of variance makes more sense in my opinion.
My data are pretty clear, untreated cells have a survival of 80% and treated cells have a survival of 5% (9). I just want to present my data correctly, but Iím really lost, itís not like my data worthless and Iím desperately seeking for statistical significance. Furthermore, t
My knowledge about statistics is pretty limited and I used graphpad prism for years and unfortunately I have no experience with other stat softwares like SPSS or R.
I hope this thread is not too confusing