2 way ANOVA and non-parametric alternatives

#1
I want to compare 4 groups: (treatment 1 tissue type 1), (treatment 1 tissue type 2), (treatment 2 tissue type 1), (treatment 2 tissue type 2).

I'm mostly interested in any difference between treatment groups, but anticipate there may be an interaction with tissue type.

Initial thoughts are to use 2-way ANOVA. Sound reasonable?

I'm not sure if the data will be normally distributed, though to complicate matters it will likely be ratio data. I could try some transformations but are there any non-parametric alternatives if these fail? I suspect i would lose the ability to measure any interaction though?

Cheers.
 
#2
Initial thoughts are to use 2-way ANOVA. Sound reasonable?
Yes, that sounds very reasonable.

I'm not sure if the data will be normally distributed, though to complicate matters it will likely be ratio data.
Even if the residuals turn out to not be normally distributed, then try a generalized linear model (a glm) (a generalisation of the general linear model - that is based on the normal distribution). A glm can have distributions like binomial Poisson, gamma, beta and many other, thus including skewed distributions.
 
#3
Even if the residuals turn out to not be normally distributed, then try a generalized linear model (a glm) (a generalisation of the general linear model - that is based on the normal distribution). A glm can have distributions like binomial Poisson, gamma, beta and many other, thus including skewed distributions.
Ah yes, i vaguely recall GLMs. I'll have to read up on it.

Would the fact that the data is a ratio make this more tricky. I was thinking initially maybe a 2-way ANCOVA, the numerator kept as the IV with the denominator a covariate.