Desktop User Guides > Reporter > Applying statistical tests > Column proportions test > Details and restrictions of the column proportions test
 
Details and restrictions of the column proportions test
The column proportions test is not suitable for all tables. When you request the test on a table that is structurally unsuitable, UNICOM Intelligence Reporter changes the specified table level (a warning message is not provided). It is up to you to make sure that the data in the table is generally suitable for testing, that the sample size is suitable, and so on.
UNICOM Intelligence Reporter displays a message if you define a column proportions test on an unsuitable table or if you change a table that has a column proportions test defined so that it is no longer suitable for the test. When this happens, you can either adjust the table so that it conforms to the restrictions described here, or you can remove the test from the table. However, sometimes UNICOM Intelligence Reporter cannot determine that a table or a section of a table is unsuitable for the test until it actually attempts to perform it; for example, when a table has only two category columns and all of the values in one of those columns are zero. When this happens, UNICOM Intelligence Reporter skips the test.
Multiplier
This test is unsuitable when more than one single multiplier is applied.
Hierarchical data
This test is unsuitable for running on lower level data when you are working with hierarchical data a hierarchical view of the data.
Grids
It is possible to run this test on grids, provided that the test is carried out on a grid table structured in the format:
rating[..].Column * rating
This test is suitable for running on grid tables in the following format when you set the level of the table to be the Top (for example, TableDoc.Table1.Level="hdata"):
rating * rating[..].Column
Weighting
This test is unsuitable for running when weighting is applied to individual columns or rows.
Rows
This test compares the proportions in each row that are formed from a variable category. The test is not performed on rows that are formed from non-category elements, such as bases and means.
Columns
For each category row, the test compares pairs of columns that are formed from variable categories, testing whether the proportion of respondents in one column is significantly different from the proportion in the other column. UNICOM Intelligence Reporter does not test columns that are formed from non-category elements or columns in which all of the values are zero. The test cannot be performed on tables that contain more than 52 category columns if you request one significance level, or 26 category columns if you request two significance levels, and it needs a minimum of two category columns.
Built-in Column Proportions tests
If a column proportions test is included in the metadata for a variable, the test is performed only if the variable is added as a single variable. If it is nested or concatenated with other variables, the built-in test is ignored to prevent possible inconsistencies in the results, and you must explicitly specify a column proportions test for the whole axis.
Nested tables
If there is nesting on the top axis, the test is performed separately for each set of columns that is formed from the categories of the innermost child variables. This means that the innermost child variables must have at least two categories. Nesting on the side axis does not change the test--each category row is always tested.
Concatenated tables
If there is concatenation on the top axis, the test is performed separately for each set of columns that is formed from the categories of a concatenated variable. Concatenation on the side axis does not change the test--each category row is always tested.
Built-in bases
If any of the variables on the top axis has more than one base, the test is performed separately for the columns formed from the categories before and after each base.
Sample size
This test relies on a large sample, which means that it might not be valid for a small sample--for example, fewer than about 30 cases. UNICOM Intelligence Reporter checks for small sample sizes, and does not carry out the test on columns with a base below 30. You can change the minimum sample size if required , by entering a new value in the Minimum Base field in the Statistics tab of the Table Properties dialog.
Multiple response variables
When there is a multiple response variable on the top axis, UNICOM Intelligence Reporter performs the overlap adjustment.
Two-tailed test
This is a two-tailed test, which means that it reports all significant differences between the proportions in all of the columns regardless of which columns contain the greater proportions.
Excluded elements
The column proportions test is not carried out for rows that have been excluded from the base using the IncludeInBase=False property. The column proportions test is carried out for columns that have been excluded from the base using the IncludeInBase=False property.
Overlap formula
Each axis can be derived from either an axis expression or an MDM variable. When an axis is derived from an axis expression, TOM will honor the MaxResponses property. When the MaxResponses value is greater or equal to 2, TOM regards the axis as overlapped. Considering that the MaxResponses default value is 2, each axis is in an overlap state by default.
When an axis is derived from an MDM variable, TOM will honor the variable's EffectiveMaxValue property. When the EffectiveMaxValue value is greater or equal to 2, TOM regards the axis as overlapped. When there are any sub-axis that are overlapped for a table's side and top, TOM regards the side or top as overlapped.
When both the side and top are overlapped for a table, and UseGridOverlapFormula is true, the grid overlap formula is applied to the table. The normal overlap formula is applied when the table's top is overlapped, otherwise the standard formula is used.
Refer to the following topics for more information regarding overlap:
TOM.IStatistics.UseGridOverlapFormula
TOM.IAxis.MaxResponses
MDM.IField.EffectiveMaxValue
See also
Column proportions test