The SPP Procedure

COVTEST Statement

  • COVTEST process-name = trend-name <trend-name, …></options>;

You use the COVTEST statement to perform covariate dependency tests that are based on an empirical distribution function (EDF). The COVTEST statement contains two essential parts: a process-name (which must be declared in a PROCESS statement that precedes the COVTEST statement) that you specify on the left side and a list of trend-names (which must be defined in a preceding TREND statement within the same PROC SPP call) that you specify on the right side. The procedure performs separate EDF tests for every trend-name that is specified in the right side of the COVTEST statement. When you include a trend on the right side of the COVTEST statement, PROC SPP performs a weighted EDF test. Performing EDF tests involves computing EDF statistics, for which the SPP procedure calculates the Kolmogorov-Smirnov D statistic by default. PROC SPP also produces an EDF plot for the Kolmogorov-Smirnov D statistic, which you can request by using the  TEST=D option. If you are interested instead in the Cramér–von Mises $W^{2}$ statistic, you need to request it via the following covtest-option:

TEST= CM | D

requests the test statistics for any type of requested weighted tests. This option applies only if you have specified a trend on the right side of the COVTEST statement. The requested weighted test statistic is applied to every trend that is specified on the right side of the COVTEST statement. You can specify the following values:

D

requests the Kolmogorov-Smirnov D statistic.

CM

requests the Cramér–von Mises $W^{2}$ statistic.

By default, TEST=D.