Dana, if I read well your message, your new variable resefc1 will have a valid value only when recefc2 fails to be given a valid value, and vice versa. Then, there would be no case with a pair of valid values in recefc1 and recefc2, and no correlation could be computed between them!! (except you thoughtfully gave, say, a zero value to recefc1 when assess=2 and a zero value to recefc2 when assess=1).

On the other hand, it is logically possible, though seemingly unlikely, that all cases with assess=1 have the same value on recefc (or all cases with assess 2 have the same value on recefc), and this would generate the message you receive.

Hector Maletta Universidad del Salvador Buenos Aires, Argentina

> Dana Ross wrote: > > Hello! > > I have a question about defining variables that are used in > test-retest correlations. > > If I have a variable that is computed under the If statement, does > that "If" condition make the new variable a constant? for instance, > i have a computed variable "resefc" that is computed as the sum of > responses 63-65. I have computed two other variables "resefc1" > (which is resefc IF assess=1) and "resefc2" (which is resefc IF > assess=2). > > in this case, does the "assess" variable act as a constant? I ask > because I am unable to do bivariate correlations on resefc1 and > resefc2; i keep getting a message that indicates "one or more of the > variables is a constant". > > any help would be appreciated. > > Dana Ross > grad student > University of Alabama at Birmingham > >