Chi-square test for independence: Options

To help you uncover patterns in the data that contribute to a significant chi-square test, the Chi-Square Test for Independence procedure displays expected frequencies and three types of residuals (deviates) that measure the difference between observed and expected frequencies. Each cell of the table can contain any combination of counts, percentages, and residuals selected. The Options dialog provides settings for uncovering data patterns.

Counts

The number of cases actually observed and the number of cases expected if the row and column variables are independent of each other. You can choose to hide counts that are less than a specified integer. Hidden values will be displayed as <N, where N is the specified integer. The specified integer must be greater than or equal to 2, although the value 0 is permitted and specifies that no counts are hidden.

Z test

Computes pairwise comparisons of column proportions and indicates which pairs of columns (for a given row) are significantly different.

Compare column properties

When enabled, significant differences are indicated in the crosstabulation table with APA-style formatting using subscript letters and are calculated at the 0.05 significance level.

Note: If this option is specified without selecting observed counts or column percentages, then observed counts are included in the crosstabulation table, with the APA-style subscript letters indicating the results of the column proportions tests.

Adjust p-values (Bonferroni method)

Pairwise comparisons of column proportions make use of the Bonferroni correction, which adjusts the observed significance level for the fact that multiple comparisons are made.

Row order display

Provides options for arranging rows in ascending/descending order.

Percentages

The percentages can add up across the rows or down the columns. The percentages of the total number of cases represented in the table (one layer) are also available.

Note: If Hide small counts is selected in the Counts group, then percentages associated with hidden counts are also hidden.

Residuals

Raw unstandardized residuals give the difference between the observed and expected values. Standardized and adjusted standardized residuals are also available.

Unstandardized

The difference between an observed value and the expected value. The expected value is the number of cases you would expect in the cell if there were no relationship between the two variables. A positive residual indicates that there are more cases in the cell than there would be if the row and column variables were independent.

Standardized

The residual divided by an estimate of its standard deviation. Standardized residuals, which are also known as Pearson residuals, have a mean of 0 and a standard deviation of 1.

Adjusted standardized

The residual for a cell (observed minus expected value) divided by an estimate of its standard error. The resulting standardized residual is expressed in standard deviation units above or below the mean.

Non-integer weights

Cell counts are normally integer values, since they represent the number of cases in each cell. But if the data file is currently weighted by a weight variable with fractional values (for example, 1.25), cell counts can also be fractional values. You can truncate or round either before or after calculating the cell counts or use fractional cell counts for both table display and statistical calculations.

Round cell counts

Case weights are used as is but the accumulated weights in the cells are rounded before computing any statistics.

Truncate cell counts

Case weights are used as is but the accumulated weights in the cells are truncated before computing any statistics.

Round case weights

Case weights are rounded before use.

Truncate case weights

Case weights are truncated before use.

No adjustments

Case weights are used as is and fractional cell counts are used. However, when Exact Statistics (available only with Sampling and Testing) are requested, the accumulated weights in the cells are either truncated or rounded before computing the Exact test statistics.

Missing values

Controls the treatment of missing values.

Exclude user missing values

Excludes cases that have user missing values for all variables involved in the test.

Include user missing values

Includes cases that have user missing values for all variables involved in the test.