All commonly used statistics on tables, such as %, Column %, and Average, are computed using the base. It is often necessary to rebase tables (i.e., modifying the sample size used in the calculation of statistics). For example, tables may be rebased to address data integrity issues or to see results among a particular sub-group. This can be done using a range of different methods, as follows.
Filtering
The most straightforward approach to rebasing a table is to create and apply a filter.
On tables where there is a desire to rebase multi-response data to the NET (e.g., rebasing to only show people that have purchased a subset of brands), you can do this by selecting the multi-response variable from the Data Sources tree and clicking Rebase Multiple Response Data in Variable(s) to NET under Modify Variable Sets in the object inspector (see How to Rebase Multiple Response Data in Variable(s) to NET for more details).
Note also that where you have multiple response data (e.g., Binary - Multi or Binary - Grid questions), it is not guaranteed that the categories not selected end up with only 0s on the table.
Modifying the Value Attributes
The base used in computations on a table can be adjusted by selecting the variable from the Data Sources tree and then clicking Data > Properties > Values in the object inspector. You can then edit the Value Attributes for that variable.
- Changing which categories have Missing Values selected (if set to Exclude from analyses, cases with that value are excluded).
- Changing which categories have a Value of NaN. This works the same way as Missing Data but is only applicable for numeric questions, and thus it is possible, for example, to have 'Don't Know' responses included in percentages selected in Statistics > Cells but excluded from the Average calculations in Statistics > Below and Statistics > Right.
A shortcut to exclude categories is to right-click on their row or column headings on a table and select Delete. This has the same effect as changing the category to Exclude from analyses in the Value Attributes.
While modifying Value Attributes is generally the preferred approach if addressing data integrity issues, it is only possible in situations where the original data file contains useful missing values. In particular, if the data file uses the same value to indicate a respondent saw but did not choose an option as is used to indicate that they did not see an option, it becomes impossible to rebase by using the value attributes.
Creating new variables that are filtered
A new version of a question can be created in which the underlying data has been modified such that certain cases are excluded. The most straightforward approach to doing this is to select a variable you wish to filter, then hover and click + > Filter > Filter One Variable Set by Another which will filter selected variables by the categories of a nominal variable. See How to Rebase One Question Based on Another Question for more details.
Deleting cases
Where there is a desire to rebase all the data in the project by removing some respondents, this is best achieved by deleting cases.