Ratio of sizes data

So I have data regarding the ratio of a cell organelle (the Golgi) to the overall cell size from 48 cells, each one belonging to one of 3 genotypes. I want to investigate whether this ratio is influenced by genotype. My first thought was simply to perform ANOVA on the data (or a Kruskal-Wallis test if ANOVA assumptions not met).

But do I need to be careful with ratio data because this essentially treats two related variables (golgi size and cell size) as a single variable. Are there more suited methods to explore this data?


TS Contributor
I'm now wondering whether ANCOVA is the way to go with this with Golgo area as the independent dependent variable and cell area as a covariate with genotype as an independent variable. Thoughts?
This was my thought after reading your first post. I Think that would be reasonable to use ANCOVA to model Golgi size (volume, area, or however you intended) as a function of genotype after accounting for the covariate cell area/volume (again whichever size measurement you planned to use).

Is there any theory or reason to believe that, for a given cell size, the relationship between golgi size and genotype changes? In other words, is it reasonable to assume that the relationship between golgi size and genotype is modified by overall cell size (or that relationship of golgi size with cell size is modified by genotype)?

EDIT: I may need to clarify-- are you trying to look at total cell size or size of the golgi apparatus? My reply assumed (possibly incorrectly)that you were interested in the golgi size. If you're interested in total cell size as the dependent variable, then switch "golgi" and "total cell" in my reply.
Thanks for the correction. Yes it is how Golgi size changes with 3 different genotypes i want to examine, but Golgi size is also related to cell size so it's the relative size i want to examine. It is unclear at this point whether the 3 genotypes influence cell size as well. I'll give ANCOVA a try but may return with more questions as I've never used it before.
I've been using this as a guide to ANCOVA in r.

Turns out that my model violates some the assumptions of ANCOVA (numbers 3 and 4 in the link). In particular the covariate, cell area, is different between some of the genotypes. The variances are also not equal.

I am a little confused because if there is a difference between cell size and not a difference in golgi size, then I would expect a difference in the ratio of golgisize/cellsize, but this is not the case.

I am unsure how to proceed.
The more statistical tests I perform the more likely I am to find one that 'looks good', which I want to avoid.