2 way ANOVA with no interaction

#1
Hi,

I have a data set with the following factors: Strain (1 depressive rat model, 2 control rat groups) and Region (8 brain regions) and receptor density (dependent variable).

I have performed Two-way ANOVA with Bonferroni correction using GraphPad.

I get a significant effect of Strain and Region but no interaction between the two factors.

My question is what I should do next. I am interested in the differences between the 3 rat groups in each region, but I am no really interested in the overall differences between regions.
There is no interaction and I have read different places on the internet that I should now do One-Way ANOVA (or pairwise t-tests).

1) Am I allowed to do One-Way ANOVA with Strain as factor and look in 1 region at a time? This way, that I know of, I will not be able to correct for the many Type I and Type II errors?

2) Should I continue with the Two-way ANOVA results that tells me in which Regions I see differences between the different rats (I only have 6 rats in each group so my statistical power is small and I almost don't get anything significant doing it this way)

Thanks for your help!
 

Karabiner

TS Contributor
#2
Strain (1 depressive rat model, 2 control rat groups) and Region (8 brain regions) and receptor density (dependent variable).
I have performed Two-way ANOVA with Bonferroni correction using GraphPad.
Seemingly, in each individual rat you took 8 measurements
(1 for each region)? In that case you'll have to do repeated-
measures ANOVA.

In addition, why or what for you performed Bonferroni-correction
is not immediately clear.

With kind regards

K.
 
#3
Thanks for your reply.

Well, the rats were decapitated and the brains were cut, but in some we were not able to get all the regions of interest and also some regions were lost due to insufficient quality of autoradiography imaging.
Therefore I have an unequal sample size; for some rats I have 8 measurements but for others, I only have 4 (or 5,6,7).
Can I still do repeated-measures ANOVA when sample sizes are so different?

Bonferroni-correction was a post-test after two-way anova to correct for multiple comparisons.
 

Karabiner

TS Contributor
#4
Can I still do repeated-measures ANOVA when sample sizes are so different?
For RM-ANOVA you need complete cases, without missings.
Alternatively, you could try or multilevel modeling. Don't
know whether generalized estimating equations (which is
easier to handle) is perhaps also appropriate.

With kind regards

K.