Hi Barry,
How about using a weighted Pearson correlation coefficient?
In SAS Proc Corr, there is a Weight option.
http://support.sas.com/documentation/cdl/en/procstat/70116/HTML/default/viewer.htm#procstat_corr_syntax07.htmI did a google search for examples of SAS Proc Corr with the weight option, but couldn't find anything.
If you can find a formula to compute the weights from the N's and standard deviations from each line in your dataset, then using Proc Corr with the weight option could be the solution.
If you can't find the formula for the weighting, I suggest calling SAS Tech Support at 919-677-8008 and ask them for an example of Proc Corr with the weighting statement and also for the formula to compute weights from N's and sigmas.
------------------------------
Brandy Sinco, BS, MA, MS
Research Associate
------------------------------
Original Message:
Sent: 09-08-2017 13:18
From: Barry DeCicco
Subject: Correlation of Scores between Two Surveys (with two N's)
Hello,
I am correlating the scores between two surveys, each across the same set of units. Each unit has one score and one N for each of the two surveys: X1, N1, X2, N2. I'm trying to calculate the correlations taking both sets of N's into account.
Does anybody have a lead to good sources and methods on this?
Thanks!
Barry DeCicco
My data would look something like this:
| Survey 1 | Survey 2 |
Unit | N1 | Score1 | N2 | Score2 |
1 | 57 | -0.460 | 18 | 0.623 |
2 | 89 | -0.324 | 51 | -0.176 |
3 | 52 | -1.544 | 90 | -1.523 |
4 | 34 | 1.694 | 75 | -0.456 |
5 | 87 | -0.314 | 611 | -0.312 |
6 | 89 | -0.053 | 28 | 0.320 |
7 | 60 | -1.624 | 698 | -2.016 |
8 | 51 | -0.500 | 321 | 0.054 |
9 | 18 | -0.896 | 42 | -0.760 |
10 | 9 | 0.552 | 87 | 1.175 |
| | | | |
I would like to get the correlation of the scores from the surveys, taking into account the sample sizes |
------------------------------
Barry DeCicco
Business Analyst
Nissan
------------------------------