I have a question, which has puzzled me for some time now. I hope this community can provide an answer. I have a method which can measure conversion of pyruvate into lactate in the body, the results are expressed in ratios, e.g. 0,2; 0,4 and so on. I have conducted several animal experiments using this method and am now interested in perfecting my animal method. Specifically, I want to know if it is better to use fasted or fed animals. I have results from a sample of fasted animals, n=15 and results from a sample of fed animals, n=5. What I want to know is, does fasting increase the variability of the results? Are the animals more "in line" with each other when they've been fed, i.e. do the results lie closer together? I cant seem to find a good way of testing for this - I've thought of using coefficient of variation or a Bartlett's or Levene's test, but am concerned as to the difference in sample size (n=15 vs. n=5). Specifically, I don't want my power to detect a difference in the variance to be diminished because of sample size-difference. Can anyone come up with a good way of testing for something like this? I'm somewhat of a novice in statistics, so please forgive me if something is unclear - or just plain wrong. Thank you in advance!
If the quantity of data you have is fixed then there is little you can do to improve your power to detect a difference. You could also use the Brown-Forsythe test which may have some slight advantage but really with five in your smaller group your prospects are not stunning.