I want to test whether the difference between right brain volume and left brain volume is significant for 3 subject groups, based on type of stroke. Each type of stroke can occur in the left or right brain, so eventually I have two independent variables, stroke type and stroke side, and two dependent variables, left brain volume and right brain volume. To look at the difference between left and right brain volume, I read that i should use a paired t-test. However, this way I can't look at differences between the stroke types or look at stroke side. Maybe this can be done by conducting multiple paired t-tests or a ANOVA? But if I would use one of these, I can only have one independent variable, so I'd have to transform my data in 6 subject groups. Also, the brain volumes of some participants are way higher than others due to weird volume measurements, but the difference between the right and brain volume is what matters. So I also don't know if a paired t-test would be the right choice for this as well. I can also use a one-way ANOVA, but then I'd have to calculate the proportion of difference between left and right brain volume first so that I'd have one dependent variable.
Hello Shakira,
Assuming that the volume measurement/quantification is consistent for all cases, then I'd suggest a 3 (stroke type) x 2 (stroke side) x 2 (hemisphere volume) anova. The first two factors are between-subjects factors, such that you'll have 6 "cells" (example: ischemic / dextral), while the last is a within-subjects or repeated measures factor. In all likelihood, your cells won't have equal numbers of cases (if for no other reason than ischemic strokes are the most common type), so you'll want to make sure you have enough cases so that individual cell values can be estimated accurately.
While the main effect of hemispheric volume is one of the tests generated in this design, from your description it sounds as if tests of interaction (e.g., stroke type x hemispheric volume) are likely to be of considerable interest for your research.
Good luck with your work.
Hello Shakira,
Assuming that the volume measurement/quantification is consistent for all cases, then I'd suggest a 3 (stroke type) x 2 (stroke side) x 2 (hemisphere volume) anova. The first two factors are between-subjects factors, such that you'll have 6 "cells" (example: ischemic / dextral), while the last is a within-subjects or repeated measures factor. In all likelihood, your cells won't have equal numbers of cases (if for no other reason than ischemic strokes are the most common type), so you'll want to make sure you have enough cases so that individual cell values can be estimated accurately.
While the main effect of hemispheric volume is one of the tests generated in this design, from your description it sounds as if tests of interaction (e.g., stroke type x hemispheric volume) are likely to be of considerable interest for your research.
Good luck with your work.
I agree entirely with David Morse my only comment is that I would write the model as a regression because I think the interactions would be more straight forward to interpret. I believe this is covered in the factorial design sections of Montgomery's Design and Analysis of Experiments. Best wishes, David Booth PS in the design phase you can deal quite easily with knowing how many observations you need to generate enough df. Another good reference might be:
https://www4.stat.ncsu.edu/~davidian/st732/notes.html
An appropriate statistical method is two way or multiple variance analysis or co-variance analysis.
- Badges
- Science topic