[b1c1jones]

I've found timings in general were lognormal a long time ago. I had to routinely apply transformations to get meaningful ANOVA and DOE results.

[hackandthink]

The obvious transformation is the log. Are there other?

[b1c1jones]

Yes, box-cox can be used, even if the lognormal itself isn't a perfect match for the actual distribution, with the correct choice of lambda.