topic Re: how do I split data into TRAINING AND TEST in BASE SAS 9.4? in Statistical Procedures

how do I split data into TRAINING AND TEST in BASE SAS 9.4?

MikeA — Mon, 19 Aug 2019 13:49:49 GMT

Hello,

I have an excel file with data for years 2017-2019. I want to import the excel file and train the 2017-18 data to predict for 2019, which is unknown. For example, if I want to predict weather college applicants would accept an offer or not based on the 2017-18 profile, how could I go about that? Can anyone please provide me with a sample code that I can use to perform such a task, by splitting the years into training and testing and how I may be able to put the variable for which I want the prediction?

Thank you,

Re: how do I split data into TRAINING AND TEST in BASE SAS 9.4?

PeterClemmensen — Mon, 19 Aug 2019 14:05:25 GMT

Hi @MikeA and welcome to the SAS Communities 🙂

A good place to start is the article Create training, validation, and test data sets in SAS

Re: how do I split data into TRAINING AND TEST in BASE SAS 9.4?

Reeza — Mon, 19 Aug 2019 14:55:32 GMT

You cannot use a standard test/train split when dealing with time series data because your splits need to be based on time. So you're looking for basic filters which can be accomplished via WHERE statements.

ie

where year in (2018, 2017);

or

where year = '2017-18';
Depending on your variable types.

Almost all procedures support a WHERE statement including a data step or proc sql.