topic Re: Export PROC PRINQUAL data in Statistical Procedures

Export PROC PRINQUAL data

micromass — Mon, 22 Jan 2018 08:57:27 GMT

Hello,

I am using PROC PRINQUAL to do some analysis on a big database of variables.

Now, since PROC PRINQUAL takes a lot of time to run on the total dataset, I have selected at random 50000 observations to run the procedure on. This generates very good output, in terms of transformations and scores.

But now I also need to generate the transformations and the scores for the observations not taken into account in the PROC PRINQUAL.

Now, if I was doing a usual PCA, this would be simple: there is an easy formula

Y_i = c_1 X_1 + ... + c_n X_n

And I could easily compute this with PROC SCORE. But since my variables now have undergone a transformation, it is not clear to me what the corresponding formula would be. Is there any way to get this formula from SAS, or to apply the PRINQUAL formula to the new observations?

Re: Export PROC PRINQUAL data

PaigeMiller — Mon, 22 Jan 2018 13:14:31 GMT

Typically, the approach to apply the model to additional data and obtain the predicted values is to include the additional data in the original data set with missing Y values. Then, PROC PRINQUAL fits the model based upon only the observations that contain non-missing Y values, but predicted Y values are produced for ALL observations.

Re: Export PROC PRINQUAL data

micromass — Wed, 24 Jan 2018 14:13:12 GMT

Thanks a lot! I read about passive observations and I tried it before. But the PRINQUAL procedure ran as long as with the additional data.

I think I'm doing something wrong. Could you perhaps give some very short sample code on how to do this?

Re: Export PROC PRINQUAL data

PaigeMiller — Wed, 24 Jan 2018 14:37:52 GMT

You show me your code and SASLOG instead of me writing some code.

By the way, what is the problem? Did it not work, or did it take too long, or something else? Your description isn't really clear about what the problem is.