J-T Subcommand (NPAR TESTS command)
NPAR TESTS /J-T=varlist BY variable(value1,value2) J-T (alias JONCKHEERE-TERPSTRA) performs the Jonckheere-Terpstra
test, which tests whether k independent
samples that are defined by a grouping variable are from the same
population. This test is particularly powerful when the k populations have a natural ordering. The
output shows the number of levels in the grouping variable; the total
number of cases; observed, standardized, mean, and standard deviation
of the test statistic; the two-tailed asymptotic significance; and,
if a /METHOD subcommand is specified,
one-tailed and two-tailed exact or Monte Carlo probabilities. This
subcommand is available only if Sampling and Testing is installed.
Syntax
- The minimum specification is a test variable, the keyword
BY, a grouping variable, and a pair of values in parentheses. - Every value in the range defined by the pair of values for the grouping variable forms a group.
- If the
/METHOD subcommand is specified, and the number of populations, k, is greater than 5, the p value is estimated by using the Monte Carlo sampling method. The exact p value is not available when k exceeds 5.
Operations
- Cases from the k groups are ranked in a single series, and the rank sum for each
group is computed. A test statistic is calculated for each variable
that is specified before
BY. - The Jonckheere-Terpstra statistic has approximately a normal distribution.
- Cases with values other than values in the range that is specified for the grouping variable are excluded.
- The direction of a one-tailed inference is indicated by the sign of the standardized test statistic.
Example
NPAR TESTS /J-T=V1 BY V2(0,4)
/METHOD=EXACT.- This example performs the Jonckheere-Terpstra test for groups that are defined by values 0 through 4 of V2. The exact p values are calculated.