The HPBIN Procedure

ODS Tables

Each table that the HPBIN procedure creates has a name associated with it. You must use this name to refer to the table in ODS statements. These table names are listed in Table 3.2.

Table 3.2: ODS Tables Produced by PROC HPBIN

Table Name

Description

Options in the PROC HPBIN Statement

BinInfo

Basic binning information and parameters

Default output

InfoValue

Information value for each variable

WOE

Mapping

Level mapping information

Default output

NObs

Number of observations read and used

WOE

PerformanceInfo

Information about the high-performance computing environment

Default output

Quantile

Estimated quantiles and extremes

COMPUTEQUANTILE

Summary

Summary statistics for the given variables

COMPUTESTATS

Trim

Trimmed statistics for the given variables

WINSOR, WINSORRATE

Winsor

Winsor statistics for the given variables

WINSOR, WINSORRATE

WOE

Weight of evidence for each bin

WOE


The following list provides more information about these tables:

BinInfo

By default, PROC HPBIN generates the Binning Information table in its ODS output. This table displays some procedure parameters and data information, which includes the binning method, the number of bins, and the number of variables.

InfoValue

PROC HPBIN generates the Information Value table if you specify the WOE option in the PROC HPBIN statement. This table provides the information value for each variable.

Mapping

By default, PROC HPBIN generates a Mapping table in its ODS output. This table provides the level mapping for the input variables. The level starts at 1 and increases to the value that you specify in the NUMBIN= option. In the mapping table, a missing value for the lower bound indicates negative infinity, and a missing value for the upper bound indicates positive infinity.

The final bin level can be less than the NUMBIN value if the input data are small or the binning variable is discrete. In this case, a warning message is printed in the log.

NObs

PROC HPBIN generates the Nobs table if you specify the WOE option. This table provides the number of observations that are read and used.

PerformanceInfo

By default, PROC HPBIN produces the Performance Information table. It displays information about the execution mode. For single-machine mode, the table displays the number of threads used. For distributed mode, the table displays the grid mode (symmetric or asymmetric), the number of compute nodes, and the number of threads per node.

Quantile

PROC HPBIN generates the Estimated Quantiles and Extremes table if you specify the COMPUTEQUANTILE option. This table contains the following quantile levels for each variable: 0% (Min), 1%, 5%, 10%, 15%, 25% (Q1), 50% (Median), 75% (Q3), 90%, 95%, 99%, and 100% (Max).

Summary

PROC HPBIN generates the Summary Statistics table if you specify the COMPUTESTATS option. This table displays the variable name, number of nonmissing observations, number of missing observations, mean, pseudo-median, standard deviation, minimum, maximum, and number of bins.

Trim

PROC HPBIN also generates the Trimmed Statistics table if you specify the WINSOR option. This table contains the trimmed minimum, maximum, mean, standard error mean, left tail, left tail percentage, right tail, and right tail percentage, and the degrees of freedom (DF).

Winsor

PROC HPBIN generates the Winsor Statistics table if you specify the WINSOR option. The Winsor Statistics table contains the Winsorized minimum, maximum, mean, standard error mean, left tail, left tail percentage, right tail, and right tail percentage, and the degrees of freedom (DF).

WOE

PROC HPBIN generates the Weight of Evidence table if you specify the WOE option in the PROC HPBIN statement. This table provides the level mapping information, binning information, weight of evidence, and information value for each bin.

When the Weight of Evidence table is printed, the Mapping table is not printed because the level mapping information is the same in both tables.

In addition to the level mapping information, the Weight of Evidence table contains some other information such as the non-event count, non-event rate, event count, event rate, weight of evidence, and information value for each bin.