It is common for researchers to want to choose a different base when analyzing Pick One – Multi questions (single response grids), Pick Any – Grid questions (multiple response grids), and other questions involving several variables. For instance, you may have a question where people rate a set of brands, but you want to calculate the ratings based only on people who told you earlier that they purchase products from those brands.
Let’s take the following example where we have a brand association question (Pick Any – Grid) which shows the proportion of people who associate a range of supermarket brands with having good quality products, a good range of products, etc. Currently, all of the results are based on the total sample of 500 people who answered this survey.
We may wish to re-base the data so that it bases the figures only on people who indicated that they have shopped at each supermarket in the past month. If the questionnaire is designed with this intent from the start, and your data supplier does a good job, then you should get the correct base automatically. However, if you change your mind about the sample during your analysis, or if the data provider has not correctly coded skips with Missing Data, then you might be left without the desired base sample. The desired bases are represented in this table:
To rebase the grid we can:
- Select Automate > Browse Online Library > Filtering > Filter One Question by Another Question
- When prompted to Select the question you want to filter, pick the Q14 – Brand Image question from the list and click OK.
- When prompted to Select the question whose categories you want to use as filters, select Q9 – Which of these have you shopped at in the past month?
- When asked if you want Q to match up the filters automatically, click Yes (more on this below)
A new copy of the question is created, where the selections from Q9 are built in as “filters”. By this, I mean that people who did not select a brand are given a missing value for that brand in the new copy of the grid. The result looks like this: