../E41084-03.epub /> ../E41084-03.mobi />

STATS_BINOMIAL_TEST

Syntax

Description of stats_binomial_test.gif follows
Description of the illustration stats_binomial_test.gif

Purpose

STATS_BINOMIAL_TEST is an exact probability test used for dichotomous variables, where only two possible values exist. It tests the difference between a sample proportion and a given proportion. The sample size in such tests is usually small.

This function takes four arguments: expr1 is the sample being examined. expr2 contains the values for which the proportion is expected to be, and p is a proportion to test against. The fourth argument is a return value of type VARCHAR2. If you omit the fourth argument, then the default is TWO_SIDED_PROB. The meaning of the return values is shown in Table 5-3.

Table 5-3 STATS_BINOMIAL Return Values

Return Value Meaning

TWO_SIDED_PROB

The probability that the given population proportion, p, could result in the observed proportion or a more extreme one.

EXACT_PROB

The probability that the given population proportion, p, could result in exactly the observed proportion.

ONE_SIDED_PROB_OR_MORE

The probability that the given population proportion, p, could result in the observed proportion or a larger one.

ONE_SIDED_PROB_OR_LESS

The probability that the given population proportion, p, could result in the observed proportion or a smaller one.


EXACT_PROB gives the probability of getting exactly proportion p. In cases where you want to test whether the proportion found in the sample is significantly different from a 50-50 split, p would normally be 0.50. If you want to test only whether the proportion is different, then use the return value TWO_SIDED_PROB. If your test is whether the proportion is more than the value of expr2, then use the return value ONE_SIDED_PROB_OR_MORE. If the test is to determine whether the proportion of expr2 is less, then use the return value ONE_SIDED_PROB_OR_LESS.

STATS_BINOMIAL_TEST Example The following example determines the probability that reality exactly matches the number of men observed under the assumption that 69% of the population is composed of men:

SELECT AVG(DECODE(cust_gender, 'M', 1, 0)) real_proportion,
       STATS_BINOMIAL_TEST
         (cust_gender, 'M', 0.68, 'EXACT_PROB') exact,
       STATS_BINOMIAL_TEST
         (cust_gender, 'M', 0.68, 'ONE_SIDED_PROB_OR_LESS') prob_or_less
  FROM sh.customers;