Combining tests and obtaining ROC curves

[Dan Taylor-5]

I have two tests, Test A and Test B

A group of 50 normals have no disease but 45 of them test negative
with Test A and 47 of them negative via Test B
Test A is said to have 45/50 or 90% specificity (true negative rate)
and Test B has 47/50 or 94% specificity.

A group of 50 patients all have unequivocally the disease. 42/50 test
positive by Test A so it is 84% sensitive (true positive)
and 44/50 test postive by Test B (88% sensitivity).

Question: I need to draw up, (using SPSS? I have the Adnaced Stas module), a
single ROC curve representing the combined test result where

Combined Specificity: Negative  = when BOTH Test A result AND Test B
result in a negative

Combined Sensitivity: Positive - when EITHER Test A OR Test B result
in a Positive


I would be very grateful for any advice on how to do this.

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Marta Garcia-Granero]

Dan Taylor wrote:

> I have two tests, Test A and Test B
>
> A group of 50 normals have no disease but 45 of them test negative
> with Test A and 47 of them negative via Test B
> Test A is said to have 45/50 or 90% specificity (true negative rate)
> and Test B has 47/50 or 94% specificity.
>
> A group of 50 patients all have unequivocally the disease. 42/50 test
> positive by Test A so it is 84% sensitive (true positive)
> and 44/50 test postive by Test B (88% sensitivity).
>
> Question: I need to draw up, (using SPSS? I have the Adnaced Stas module), a
> single ROC curve representing the combined test result where
>
> Combined Specificity: Negative  = when BOTH Test A result AND Test B
> result in a negative
>
> Combined Sensitivity: Positive - when EITHER Test A OR Test B result
> in a Positive
>
>

ROC curves are used for QUANTITATIVE tests, while you have binary tests
(negative/positive). You can't use ROC curves. Create a composite binary
variable with value=0 if both tests are negative, and value=1 if either
is positive, and run a contingency analysis (ask for row and column
percentages).

HTH,
Marta García-Granero

--
For miscellaneous statistical stuff, visit:
http://gjyp.nl/marta/

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD