I have two sets of histology data for the same set of patients, each with two samples. Both are made based on fraction stained cells relative to total number of cells (stained+unstained), but their ranges are quite different:

Histology parameter A

Range: 0 - 0.8,

Median value: 0.3

Histology parameter B

Range: 0.01 - 0.15

Median value: 0.02

I want to compare the overall variation for these two datasets and I'm wondering if I can do this by by calculating the total sum of squares (within mean square + between mean square, ANOVA).

Doing this gives me the following results

Histology parameter A: 0.06

Histology parameter B: 0.0006

Since the histology parameters have very different ranges, I thought it would be better to adjust both of them to be relative to their maximum value, giving me the following :

Histology parameter A

Range: 0 - 1,

Median value: 0.4

Total sum of squares: 0.08

Histology parameter B

Range: 0.03 - 1.0

Median value: 0.15

Total sum of squares: 0.03

My conclusion from this would be that there is a larger total variation in histology parameter A than in B.

I'm not a statistician and I'm therefore wondering if this makes sense and if it is something I'm "allowed" to do.

Alternatively, I would have to calculate coefficient of variation (CV), giving me 0.69 for A and 0.88 for B. This results in an opposite conclusion (B shows larger variation tham A), and I'm therefore not sure if interpret my results correct.

Any help would be highly appreciated! :)

The problem might seem odd to someone, but I'm also comparing the within patient variation with the between patient variation using the intraclass correlation coefficient (ICC), so any advices in this direction is not needed :)


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.