By default in Displayr, NETs are excluded from Column comparisons. This article describes how to include the overall NET column when performing column comparison stat testing. To test against sub-nets or other specific columns, see How to Specify Columns to be Compared in a Table. This article explains how to go from a table where the NET column is not tested:
To one that tests the overall NET column against the other categories in the columns:
Requirements
A table with:
- Compare columns selected in the Appearance > Significance menu.
- A main (overall) NET or Total column, see Banners for more detail on how to create these.
- A categorical (nominal) question or banner with at least one categorical variable set in the columns (cannot be a date question).
Method
- Select one of the column headers in the table and go to the object inspector.
- There go to Appearance > Significance > Advanced. IMPORTANT: Please read through How to Apply Significance Testing in Displayr to understand what each function performs before making updates to the statistical assumption settings in your project.
- In the Test Type tab it is also recommended that you:
- Change the Proportions setting to Survey Reporter or Quantum.
- Change the Means setting to Survey Reporter or Quantum.
- On the Column Comparisons tab, uncheck Recycle column letters.
- In the Overlaps field, select Independent or Dependent as desired, see Overlaps for more info.
- There is a Compare section for the variable set at the bottom. (If it is not there, ensure you have a column header selected before opening the Advanced settings.) Check Include NETs on the appropriate option selected, or specify which groups of columns should be compared in one of the Custom fields. See How to Specify Columns to be Compared in a Table for more on this.
Additional changes you may also consider:
Change Weights and significance to Kish's approximation in the Significance levels section.
Un-tick Within row and span under Column comparisons. This will determine the multiple comparison correction so that the number of tests used in the calculation is based on the entire table and not just within the span, see How to Apply Multiple Comparison Correction to Statistical Significance Testing.