When you run PROC HPSPLIT, the following set of symptoms might occur:
- The "Number of Observations Read" value that is displayed in the SAS output is smaller than the "Observations Read from the Data Set" value that is displayed in the SAS log.
- The procedure uses fewer observations than expected, even when there are no missing values.
- When scoring the DATA= data set, the SCORE OUT= data set contains fewer observations than the DATA= data set.
To work around the problem, use a PERFORMANCE statement with the NTHREADS=1 option:
performance nthreads=1;
Here is an example:
proc hpsplit data=Wine seed=15533;
class Cultivar;
model Cultivar = Alcohol Malic Ash Alkan Mg TotPhen Flav
NFPhen Cyanins Color Hue ODRatio Proline;
grow entropy;
prune costcomplexity;
performance nthreads=1;
run;