I have a dataset which has a predictor (Temperature), a response (Developmental Time) and two separate factors Species ( A & B) and Population (High, Low, Middle). I want to compare whether the populations and the species differ from each other (is A different from B or is low A different to high A, but not necessarily together (e.g., Middle A is different to Low B). Theory suggests the data should be non-linear and follow a power function, and indeed, it's neither linear or normally distributed. I can transform both the predictor and the response by logs to correct this, and follow up with an ANCOVA (DT~Temp*Pop, dataset=dataset), but I'm unsure whether this is the most powerful method possible. I've tried using boxcox transformations to simply transform the response, but to no avail. In short, I want to compare different non-linear models to each other, but have no idea how. I would also like to know how to draw this non-linear model through a plot (for High A, Middle A, Low A, High B & Low B) Thanks, D [[alternative HTML version deleted]]