topic Re: How to denote different categories using proc corr? in Statistical Procedures

How to denote different categories using proc corr?

cdunlea — Tue, 08 Sep 2020 21:32:40 GMT

Hi all,

I would like to perform a Spearman correlation on the scores from a single variable in my dataset. I have six other variables that categorize all participants into 6 different groups by giving them a 1 or a 0.

How can I use proc corr to say that I want to calculate the correlation coefficient on the scores for each different group as defined by the 6 other variables? An example table of what I am hoping for is below.

	Group 1	Group 2	Group 3	Group 4	Group 5	Group 6
Group 1		Correlation coefficient for score on test
Group 2
Group 3
Group 4
Group 5
Group 6

Re: How to denote different categories using proc corr?

PaigeMiller — Tue, 08 Sep 2020 21:35:01 GMT

Please show us a portion of the original data

Re: How to denote different categories using proc corr?

cdunlea — Tue, 08 Sep 2020 21:52:43 GMT

Here is part of my dataset with the variables of interest.

Obs Score group1 group2 group3 group4 group5 group612345678

71.00	1	.	.	1	.	.
20.00	.	.	.	.	.	1
29.00	.	.	.	.	.	1
44.00	.	1	.	1	.	.
20.00	.	1	.	1	.	.
63.00	1	.	.	1	.	.
50.00	.	1	.	.	.	.
62.00	.	1	.	1	.	1

Re: How to denote different categories using proc corr?

cdunlea — Tue, 08 Sep 2020 21:58:01 GMT

Here is the dataset with no formatting errors.

Score	Group1	Group2	Group3	Group4	Group5	Group6
71.00	1	.	.	1	.	.
20.00	.	.	.	.	.	1
29.00	.	.	.	.	.	1
44.00	.	1	.	1	.	.
20.00	.	1	.	1	.	.
63.00	1	.	.	1	.	.
50.00	.	1	.	.	.	.
62.00	.	1	.	1	.	1

Re: How to denote different categories using proc corr?

Reeza — Tue, 08 Sep 2020 22:45:26 GMT

Usually observations have to line up somehow....how do they align here? How do I know which group 4 obs to match with group 1, is it only when the observation has 1s in Group 4 and Group 1? But that doesn't make sense because you only have one value for that observation so how does correlation get calculated at all here?

Re: How to denote different categories using proc corr?

PaigeMiller — Tue, 08 Sep 2020 23:31:52 GMT

I have to admit I don't understand. A Spearman correlation cannot be calculated if the only values in the data are 1 or missing. Even if (as you said in your original message, but not in your most recent messge) that the values are 1 and 0, I don't see this as a candidate for a Spearman correlation. Correlation between two binary variables is measured by the Phi Coefficient, which is an output from PROC FREQ.

I'm also mystified by your title, "how to denote different categories", which doesn't seem to relate at all to anything you have written.