Hi, I'm analyzing behavioural mice data, such as locomotivity, anxiety and learning and memory. by qq plot I confermed that the data I collected have a considerable variability. I runned normality test and confirm this variability. But I have a large number of animals by group: 3 groups; n=25, 22, 23. So I have two questions regarding wiht. is animal number larger enought to consider a normal distribution?, and if yes, there is another observation: those animals are outbred, and I have riden an article that mentioned that outbred mice tend to have more variability in behaviour. so in the case I can run a parametrical test, Do I have to be present the mouse strain?. I really appreciate your suggestions
If your interest has more to do with the center parts of the distributions, then it is probably not critically important whether or not they conform to a Normal Distribution. But if you are interested in the tails of the distributions, your conclusions may be on shaky ground and you may need to use small sample statistics and/or nonparametric models.
Although I do not have a direct answer to your question on the normal distribution, if you are doing parametric group comparisons, you could calculate the effect size (such as partial eta squared) and conduct a power analysis using Cohen's d (the e-book, power analysis for the bevhavioral sciences, is accessible through google scholar), to determine the number of subjects required to obtain a statistically significant effect. There are also many online calculators to determine sample size given your effect size and power estimates. Here's a site that has an overview of power and sample size calculators:
http://powerandsamplesize.com/Calculators/
Hope this helps,
BD