04-19-2017 08:47 PM - edited 04-19-2017 09:01 PM
I've conducted many Proc Univariate runs on data, each run creating a seperate small dataset.
Dependent Variables (totalling around 200) are discrete, a rank of: 0, 1, 2, 3, 4, 5, 6
Independent Variable(s): continuous
Example of SAS code for a single run:
proc univariate data=SAS_1.combined; var i_50503; class rg5_21705 rg5_21502 ; ods output moments=SAS_1.i_50503__rg5_21705__rg5_21502 ; run;
In plain English, my thinking is:
50503 is a function of 21705-level and 21502-level
Each small resulting dataset contains the Mean for the particular run. As well as lots of other 'moments' numbers.
I'd like to compare the Means in the different dataset files.
Is there a way for SAS to do this?
Or must I extract the individual Means from each dataset, and put into a new Just-Means dataset?
Might there be an alternative to Proc Univariate that can look at all possible combinations of the dependent variables levels, along with Mean score for the independent variable?
If further information is wanted I'll be happy to provide it.
04-24-2017 11:08 AM
Could you just query dictionary.tables and dictionary.columns to get all table and column names in your library, and write a macro to run your comparison using the full list of tables and columns, and then combine all result sets into a single dataset? This is what I do.
04-24-2017 06:03 PM
Please elaborate. I have 400,000 datasets. Seems absurd to append them all into one enormous dataset.
Would be far, far better to ask of SAS, "Which of these has the largest xyz data value?"
Hoping for a better way.