I want to evaluate whether some drugs increase lifespan of a Drosophila disease model. However I am confuse of which is the best statistical test for this experiment. Is Kaplan-Meier a good method, if so which cut off do you recommend me to use? Or is it better to use other methods such as Gompertz or Cox regression?
Ask
Definitely survival analyses are tools of choice. Kaplan Meier is rather crude and OK for preliminary analyses. Cox regression could be used to compare the effect of various treatments on lifespan and understand the role of covariates. If you want to estimate lifespan directly it might be better to go for parametric modeling using an explicit model of the hazard rate (Gompertz is one of them)
Thanks Frederic, I read a lot more and I think I understand the analysis uses and limitations. The only think that still puzzles me is that I had 3 repeats for each treatment (2 different treatments, with 3 vials each repeat, inside each vial I have 20 flies). At the beginning I thought to make an average of the Km survival of the three repeats per treatment. Plot it with error bars and do the statistical analysis. But know I am thinking I could also treat each fly as an independent subject (as long as I have the same treatment treatment) regardless the vial they were in, as this makes more sense for using the survival analysis. My rational to do this is to think as if this was a Phase III clinical trial performed in different countries, but I don't know if this makes any sense at all.