Hello everyone,
I'm currently trying to find out which statistical test I can use to show differences between two or more populations.
I had cells, which I infected with a virus +/- additional treatment with an inhibitor. Then I analyzed infection rate by FACS and calculated the mean percentage of infection.
Now, I want to show that the inhibition effect is significant. I found some publications where a two-sided t-test was done. However, I'm not sure if viral infection really follows normal distribution? Could I use Wilcoxon Rank Sum test instead? Or is there anything else that is suitable?
Thanks for the help in this for me quite confusing area. :)
Best
You need to give us little bit more information. What is the sample size? Have you check for normality? The same cells were infected with + and then - right? Are you adjusting for any variable?
If it is the same cells infected, your sample size is large (about 50) and the means are normally distributed (I suggest you check for normality) then you can run a paired samples t-test. If sample size is small (less than 30) then you can use Wilcoxon Rank Sum Test. Provide more information on the questions I raised and I can tell you more...
It's unclear to me what your unit of observation was, so I don't quite understand what the dependent variable is. If the units of observation are each cell, then the outcome is "infected"/"not infected", which is a categorical variable and could be analyzed with a chi-squared test or logistic regression. If the units of observation are multiple petri dishes or something like that, each with a bunch of cells (e.g., if you had several cultures that got the treatment and several that didn't, and for each one you got the percentage of infected cells in that culture), then your outcome variable is basically continuous but it may or may not be normally distributed, and you'd have to consider the stuff that Eric suggested above. If you just had two observations (i.e., one sample of cells that got the treatment and one that didn't, and the corresponding percentage of infection in each sample) then there's no statistical test that could be done as far as I know.
So I had about 200000 cells per well which I treated in triplicates with 4 different concentratrations of inhibitor or no inhibitor at the same time (so in total 4*3+3 =15 samples with 200000 cells each). Then I analyzed the infection of 10000 cells of each of the 15 samples, and calculated the mean. By FACS analysis I just get the percentage of cells that show a higher fluorescence (indicating infection) than non-infected cells. How can I test this for normal distribution? Is the sample size the number of counted cells or the triplicates?
It sounds like you have 5 conditions with 3 observations each. (presumably you have 15 numbers, each a percentage of infected cells, for your 15 samples.) That doesn't sound like enough to do meaningful statistics, but maybe there are specific methods that have been developed for this line of research. What is done in similar articles in your field of research?
So: Y = proportion of infected cells, and your independend variable is X = dose.
Sounds like a dose-response experiment. The standard analysis of dose-response curves should be appropriate here (Google!). However, the small sample size and the small number of tested doses will likely not give you very precise estimates of the potency and efficiacy of your inhibitor.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA