Joining Surveys
'Joining' is the process of connecting cases from different surveys together to form a longer case. There are many situations where a requirement to join surveys arises:
- Any survey, which refers to relatively static data held externally. For example, in a survey of stores in a chain, it would be possible to have details of store size, location and manager etc. held in a survey which is referenced by a "mystery shopper" survey.
- Any form of "Before and After" survey where individual cases are important and can be identified. For example, a study of clinical or medical treatments where observations are taken before, during and after a course of treatment.
The essential requirement is that the same respondents (or, more generally, observation units) are present in both source surveys and that each has a key variable with unique values.In the example we have conducted a survey of a panel of respondents and now we want to connect responses for each case with demographic details of the corresponding panel member. Once the connection has been made it will be possible to cross-analyse data from both source surveys.Joining surveys is mostly simply done by using Snap's File | Join Import operation. Beginning by identifying the first source survey, the program will proceed to ask for details of the second, and for the key variables in each to be identified. All variables from the first survey will appear in the result but it is possible to selectively include or exclude individual variables of the second survey.
One issue with joining surveys is that it is likely that there are name duplication problems. For example, the first survey may have a Q3 "When did you travel?" and the second may have Q3 "What age are you?". Clearly they are different variables, so in order to distinguish between them Snap requests a name pre-qualifier. The default is "R." which, if chosen in the above context, would result in questions Q3 "When did you travel?" and R.Q3 "What age are you?" appearing in the final joined survey.
The example uses survey snHealth Club 2003 along with snHealth Club Members.
Step By Step
Step 1
At the Survey Overview window highlight survey snHealth Club 2003 and choose the File | Join Import option. The Join Import dialogue that appears suggests that you clone the original survey first so do that (if you do not, the membership data will be imported into the original survey and is a difficult process to undo if you need to redo it).
Step 2
At the next page, choose the snHealth Club Members survey as the "survey to import from".
Step 3
On the following page identify the key variables for each survey. This is the variable that identifies corresponding cases in each survey. It should be unique at least in the second survey (snHealth Members in our example) but there is no check on this. When selected note that both variables are of the same response type.
Step 4
The next page requests details of which variables are to be imported from the second survey (all of the variables of the first survey will automatically be included). Select all of the question variables – in the general case, it would be sensible to include all non-Note variables. The option to "Include sources" may be used where you are selectively importing derived variables and you want them to remain viable after the transfer (but see the caveat below).
Step 5
The next page asks for the prefix to be applied to names of the second survey as discussed in the Introduction, above. There is also a setting to specify whether all or matching cases from the second survey are to be imported. By specifying matching cases then only those cases referenced in the initial survey will be imported. If all is selected then all matching cases will be imported followed by new, near empty cases for the others – effectively then, new cases added in that way would show No Reply for the subject survey questions.
Step 6
The penultimate page reviews the variable names that would be generated. There is a possibility that the import process can generate illegal names so an option is provided to generate new names but any generated in this way are abstract and difficult to connect to any meaning in the context of the survey.
Step 7
On the last page a review is given of all the options selected. Click Finish and the join import will be performed.
Step 8
The final step is to take a few sample cases and verify that the join has been successful.
Snap's Join operation joins two source surveys together. Where there are more than two surveys, it is necessary to join previously joined surveys in combination. For example, given three surveys A, B and C, first join A and B together (to create AB), then join that with C to create ABC. The second join operation does not necessarily share key variables with the first.
Note: Where derived variables have been transferred during the Join process, it will be necessary to redefine them in terms of the newly allocated variable names. For example if a derived variable D1 uses Q1 in the reference survey, and the standard name prefix R. is selected, all the references to Q1 will need changing to R.Q1.