Merging Similar Surveys
The requirement to merge surveys arises in many different situations:
- Surveys run on an annual basis. It may be of interest to analyse trends through two or more years. It would be fairly common for the questionnaire to change slightly from year to year so those alterations must be taken into account.
- Surveys on one subject but with a variety of different focal points, for instance one survey of users and another of non-users of a particular service.
- Surveys conducted for different locations or sites. For example a survey of similar tourist attractions which varies between sites because of the specific facilities available at each site.
- Surveys conducted using two different interviewing methods. For example, a survey conducted by paper and by PDA.
Merging two identically structured surveys using Snap is very simple. This article will demonstrate merging two surveys that differ in composition. It assumes that a survey has been conducted of Health Club users over two consecutive years. Amendments were made to the first year's questionnaire to that used for the second year.
The main differences between the Health Club surveys for 2003 and 2004 are at Q4. The 2004, survey has one code (Jacuzzi) removed and two codes (Solarium and Sauna) added by comparison to 2003. There is also a difference at Q6 where an extra question has been added for the 2003 survey to enquire about parking and access.
Step By Step
Step 1
Start by cloning the 2004 survey (perhaps calling it snHealth Club 03-04). Do not clone the raw data.
Step 2
Edit and add variables to the new survey, so that it includes everything that is in both source surveys. In our example, that means we need to add a 'Jacuzzi' category to Q4. It can be added anywhere in the list but to simplify the explanations, insert it as code 4 (between Cycling machine and Solarium).
Step 3
Open the 2003 survey and export all the data as a new CSV file from the data entry window. Be sure to check the Expand multiples, Header row and Code labels options.
Step 4
Open the CSV file (in Excel) and adjust the headings of the columns for Q4 to match the merged survey. That is, the heading for 'Swimming pool' should be Q4:7, 'Restaurant' should be Q4:8 and 'Bar' should be Q4:9. Add in new columns for Solarium as Q4:5 and Sauna as Q4:6.
Step 5
Do the same with the 2004 survey (export and edit in Excel). Change each of the codes from 'Solarium' onwards to be one code higher than they were when exported and add a new column for Jacuzzi as Q4:4.
Step 6
Open the merged survey (snHealth Club 03-04). In the data entry window, import each of the two files created above in chronological order. Be sure that the Expand multiples, Header row and Code labels options are checked.
Step 7
Check some sample counts to verify that the merge is correct.
The example given above demonstrates merging two surveys that have data from two different years. If required a new variable can be created to compare one year against the other. To do this, create a new derived, single response variable called YEAR. Set the label for code 1 to be '2003' and the value to be 'case<=121' (there were 121 cases in the 2003 survey). Set the label for code 2 to be '2004' and the value to be 'true', this will capture all remaining data. Now it is possible to perform analysis by year.
Where the two surveys relate aspects other that the year of study, for example where they relate to different locations or locations, employee departments etc., make up similar derived single variables using the same principles.
Where there are more than two surveys, merge previously merged surveys in combination. For example, given three surveys A, B and C, first merge A and B together (to create AB), then merge that with C to create ABC