The HPPLS Procedure

All the general factor extraction methods that are available in PROC PLS are supported by PROC HPPLS.

The RLGW algorithm, which is available in PROC PLS to compute extracted PLS factors, is not supported by PROC HPPLS.

PROC PLS can specify various methods to be used for cross validation. PROC HPPLS supports test set validation only by using the PARTITION statement.

The CLASS statement in PROC HPPLS permits two parameterizations: the GLM-type parameterization and a reference parameterization. The HPPLS procedure does not mix parameterizations across the variables in the CLASS statement. In other words, all classification variables are in the same parameterization, and this parameterization is either the GLM or reference parameterization. In PROC PLS, only the GLM-type parameterization is supported.

The HPPLS procedure does not support the EFFECT statement, the MISSING= option, the VARSCALE option, and the PLOTS option that are available in PROC PLS.

The syntax of the OUTPUT statement in the HPPLS procedure is different from the syntax of the OUTPUT statement in PROC PLS. In the HPPLS procedure, you do not need to provide a prefix in the OUTPUT statement. A default prefix is used if you do not provide one. If you do not specify any output statistics in the OUTPUT statement in PROC HPPLS, the output data set includes the predicted values for response variables. Furthermore, although the OUTPUT statement in the PLS procedure includes the input and BY variables in the output data by default, PROC HPPLS does not include them by default so that it can avoid data duplication for large data sets. In order to include any input or BY variables in the output data set, you must list these variables in the ID statement.

The HPPLS procedure is primarily designed to operate in the high-performance distributed environment for large-data tasks. By default, PROC HPPLS performs computations on multiple threads. The PLS procedure executes on a single thread.