Some one on SAS-L asked how to split a SAS data se...

2009-06-14T12:13:57.869-05:00

Some one on SAS-L asked how to split a SAS data set into N pieces given a N-by-1 probability vector, while the distribution of some given variables, TDSP and VS here, should be the same in the smaller files which equals the distribution of the original file.

To ensure randomness, I used K/N algorithm which built upon conditional probability of spliting. SAS-L had some extensive discussion on this methed several years.

What I did here is extending the K/N algorithm to take care of control variables. Taking any value combination of control variables as a strata, the randomness is archieved within this strata since the distribution of control variables are given by the original data set.

Hi, I was wondering whether you could elaborate o...

2009-06-14T07:03:54.170-05:00

Hi,

I was wondering whether you could elaborate on what this code is intended to accomplish.

I gather there are two inputs. One is a probability vector, which I take to be a vector of non-negative weights which sum to one. The other input is a SAS table.

What I don't understand is what the desired outcome is. What does it mean to split a table according to a vector?

For instance, if the weights were .1, .2, .3, and .4, and the table consists of 1000 records each containing a single integer between 1 and 1000, no duplicates, what is the expected outcome?

Thank you.

Comments on SAS Programming for Data Mining Applications: Randomly Split SAS table exactly according to a gi...

Some one on SAS-L asked how to split a SAS data se...

Hi, I was wondering whether you could elaborate o...