Net difference test

The net difference test deals with each row independently and compares the difference between the proportions in one pair of columns with the difference between the proportions in another pair of columns. The result is tested for significance at the selected level. The results of the test are displayed in a separate column.

For example, if the difference between the proportion of women preferring Brand A before and after testing is larger than those preferring Brand B, and the difference between those two proportions is significant at the selected level, the result is flagged as significant and the letter S is displayed in the net difference column of the row for women.

When you specify a net difference test, to specify which columns to test and where to place the result, you must also add a ntd special element to your table, immediately after the four columns you want to test. If the table is nested, this must be in the innermost nesting level.

When you specify a net difference test, to specify which columns to test and where to place the result, you must also add a net difference item to the variable. If the table is nested, this must be in the innermost nesting level.

UNICOM Intelligence Reporter carries out the test on the four columns preceding the net difference item.

The null hypothesis is that the two figures being compared are equal, that is, the difference between them is zero. The following table shows the symbols that are displayed on the table for different results.

Result	Displayed
Significant at the selected level	S
Significant at the lower selected level but not the higher level	s
Not significant at the selected levels, but significant at the 32% level	NS
Not significant at the 32% level	E

▪If the result of the test is significant at the specified significance level, places the letter S in the additional column.

▪If you run the test at two significance levels, results that are significant at the higher level are displayed using an uppercase S, and results that are significant at the lower level are displayed using a lowercase s.

▪If the result of the test is not significant at the selected levels, but is significant at the 32% significance level (equivalent to a p value of 0.32 or a confidence level of 68%, indicating that the difference between the figures being compared is at least one standard deviation away from zero) places the letters NS in the additional column.

▪If the difference between the figures being compared is not significant at the 32% significance level (that is, the difference is less than one standard deviation away from zero) places the letter E in the additional column.