As far as I know, CFU counts from bacterial suspensions should approximately follow a Poisson distribution. So in order to perform a One-way Anova, a mathematical transformation to achieve normality should be done. Am I correct on this? And which transformation would be ideal in this case? .

My issue is to compare means of colony forming unit counts from 3 groups of 8 mice. Each group has a different genotype. Shapiro-Wilks test for normality showed that the counts do not follow a normal distribution.

Thank you very much.

At least some of my counts are very large (mean counts for the three different groups are: G1: 337150; G2: 43000, G3: 16536000. However, I am unsure wether Poisson regression may be performed on counts this large.

The variance of the samples is huge (2.28397E+10^11; 4102357143; 9.29586E*10^14). This would constitute over dispersion, am I correct?


There is no way to transform count data to a true normal distribution. Counts only have integer, non-negative values, while values from a normal distribution may be any real number.

What you need to do is Poisson regression, as discussed also on this page.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.