Corrleation Output code "a"

[lizo]

What does it mean when your SPSS Correlation output has an "a" and note says cannot be computed because at least one of the variables in constant?

Thanks!

[Mahalabobis]

It means that there is no variance in one of your variables, e.g. the values for that variable are the same for every case, hence the case values for the variables are "constant".

I believe it is also possible that there is no data in one of the variables as well.

G.

[hansolavvelle]

I am working on a intertester-reliability project, and I am currently facing the same kind of problem. Tester #1 gets the same score on each patient, and it isn´t possible to calculate kappa due to "cannot be computed because at least one of the variables is a constant". But the percentage agreement is neiter 0 nor 1.. How do I solve this?

[lorb]

A constant has no correlation with any other variable because when all the values are the same it can't possibly tell you anything about another variable. Maybe it helps if you think of correlation as a number that tells you how well the small values of one variable predict the small values in another variable, and the same for big ones. Something that you might want to do (depending on how your data looks like and what you want to know) is a T-Test that tells you whether the mean of the moving variable is significantly different from the constant.

[hansolavvelle]

Thank you ! I cannot use a T-test because my data is dichotomous, just merely pain (1)/no pain (0). What would you suggest is the best method to find agreement in the following data in SPSS (eg.):

Tester #1 Tester #2
0 0
1 0
0 0
0 0
1 0
0 0

[lorb]

Difference of proportions comes to mind.
edit: some keywords to google/look up are fishers exact test and kappa test for inter rater reliability