Incorrect inset statistics for weighted comparative histogram


If you generate a comparative histogram in PROC CAPABILITY with the COMPHISTOGRAM and WEIGHT statements, the inset will incorrectly contain the unweighted statistic for any of the following keywords specified in an INSET statement:

        SUMWGT          VAR
     MEAN            USS
     SUM             CSS
     STD             CV

Percentile statistics specified in an INSET statement using the following keywords will produce incorrect percentiles in the inset:

P1              Q3
P5              P90
P10             P95
Q1              P99
MEDIAN          QRANGE

Finally, the NOBS keyword in an INSET statement will always display 0 instead of the total number of missing and nonmissing observations.

There is no circumvention for this problem. The correct weighted statistics can be obtained from a corresponding analysis in PROC CAPABILITY with the CLASS= variable specified on a BY statement, or from PROC UNIVARIATE with the CLASS= variable specified on a BY or CLASS statement.