I am doing a test using fibers that receive the treatments I am trying to analyse.
I test 40-60 fibers and split each fiber in 3 segments, 2 treatments and one untreated control. My data is very skewed so i run a box-cox transformation then I run an ANOVA using the fiberID as blocking variable. I think I am confident with the method per se and I am able to tell differences between treatments and against the control. However, the people I need to present to, are used to see these results as percent change against the control, these fibers are heterogeneous and the data is very skewed so if I calculate percent change of each fiber, my standard deviations are very large so it is difficult to see how these groups can be different, is there a way I can present this type of analysis in a way that is as intuitive and easy to understand as "treatment x is y% higher than control"?.
Thanks,
A
One idea is to plot the paired data (control vs. treatment 1).
In this plot, points above or to the left of the 1-to-1 line are higher After than Before.
http://rcompanion.org/handbook/images/image118.png
A bar plot of differences may be easier to understand for some:
http://rcompanion.org/handbook/images/image119.png
Source:
http://rcompanion.org/handbook/I_04.html
It sounds like you're doing a within subject design comparing the treatments to the control. Would your audience relate to a statement like 'in X% of the fibers the treatment was higher than the control'? Then plot the differences in a histogram (this would show the skewness of the data and might help them understand why the % average might not be the best of metrics).
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis