tests of agreement between diagnostic tests

#1
I need to assess agreement between to variables. The problem is that one of the variable is a continuous variable and the other is a categorical variable. I create a test that gives a certain value to a the compression of the brain. The gold standard test gives a categorical value among compressed, non compressed and partially compressed. I don´t know what test could help to probe an agreement between theses variables. I´ve been told to use a Balnd-altman plots for my purpose. What do you think?
 

hlsmith

Omega Contributor
#2
So you have a screening diagnostic (continuous) variable and a binary gold-standard, correct. What types of values does the continuous predictor provide? If it provides predictive probabilities why not use those. I can think of multiple options, but I think you need to provide some more information first.


Thanks.
 
#3
Thanks for you answer. Let me explain this problem better.

As I said before, brain compression is classically classified among "Compressed" - "Non Compressed" and "Partially Compressed"

We generate a compression ratio that gives a numerical value for the compression. This ratio takes values between 0 and 1 meaning 0 is absolutely compressed and 1 means absolutely non compressed (in fact this value es not possible because it would mean that there's no brain in this area).

I have to show that there is a correlation between these variables.

Does this make sense to you?

Thanks
 

hlsmith

Omega Contributor
#4
You didn't say correlated before. Check out the "Point-Biserial Correlation Coefficient".


If not check out:




Spearman rank-order
Phi
Tetrachoric

Biserial
Gamma