insufficient memory

[cecily]

My code is just:

proc logistic data=freq, Plots(Label Maxpoints=None)=all descending;

model tag= video1-video30 three1-three30 old1-old30 yong1-yong30 get1-get30/INFLUENCE;

run;

My laptop has 8 GB memory 256SSD.

But still insufficient memory. How to solve this problem???

My file contains 120000 observations , 150 predictor  variables.

[Kurt_Bremser]

What is your SAS environment?

- SAS UE?

- local SAS installed?

- remote SAS with Enterprise Guide or SAS Studio?

[cecily]

My environment is windows8.

Yes . Local SAS installed.

I am using sas9.4

[ballardw]

I would try this without the plots to see if that helps as something very easy to try.

[cecily]

I would like to see the Plots to find the influential points. Without the plots.....I think this code is useless

[PaigeMiller]

Logistic on that many predictor variables not only causes memory problems, but it also would cause interpretation problems if you were able to actually run the model — in the sense that any resulting model would be extremely difficult to use or interpret because of the correlation between the predictors.

An alternative is to use PROC PLS, which has the advantage of using an algorithm that is not as susceptible to memory issues as is PROC LOGISTIC, but it also provides more interpretable models in the case of correlation between the predictors. Win-win!

[cecily]

my main purpose is to do the classification. Could the Proc PLS be used to do the classification?

[PaigeMiller]

"Classification" is a broad term, but yes, you could use the PROC PLS results to do some types of classification.