I have done numerous life span trials on NGM plates spread with OP-50 and do find some variability, depending upon the batch of plates and the temp in the lab (the plates are stored at 20, but the lab temp varies). Therefore, I always run a plate of N2s in parallel with the mutants. On the other hand, the variability is minor enough so that if I am testing a mutant that is ‘really’ different from N2, each experiment will show a clear difference between the N2 and mutant. And I have not had a problem finding a statistically significant difference if I just sum a number of experiments run in parallel. Hope this helps.
i have a little bit questions on how to average the data obtained from trials…
i am running some survival assays, and sometimes, the results showed huge variabilities (somehow, the trend was there, i.e.: what positive result remains positive, just that the figures are changing)… which gave me a headache then, and it even came out with the idea whether to drop the “outlier” result set from all the trials, thus to minimize the standard deviation :-[
I would second the comment on being hesitant to remove “outliers.” I also sometimes have worms that live “too long” (whatever that means) and make the data look strange. But, the long living worms are real and represent real phenotypic variability. Since I do not want to 'throw out" any data unless there is an explicit reason (e.g, I accidentally transferred a larva with my adult or …?), I plot the summed data for all of the experiments of a particular mutant. This makes any individual experimental bump less significant. I also make sure that the mutant longevity experiments are all compared to similar N2 controls done in parallel (to control for plate and lab temp and food ‘quality’ variables). To do statistics comparing multiple trials with minimization of the effects of outliers, you can do two sets of statistics - one on the pooled life expectancy data and one using the MEDIAN life span for each strain. That removes any large effect of outliers.