Rotator Survey Quality Control Checklist

After compiling a survey created with Rotator Surveys Model Builder, the system adjusts the model and makes corrections automatically. However, in order to guarantee that the survey is perfect and ready to be applied, please execute the following quality control activities:

1. Check integrity of question wordings of your Survey
  • Check question by question and verify that the short texts and keywords are consistent with the question text. If there are incongruences between them, there will be confusion during the analysis.
  • If you have duplicated any question, please verify that the new question has a unique short text for its consistency.

2. Apply marks to questions when needed

Remember that typical choices such as �Don�t know/Not answer�, �Does not apply�, �None of the above�, �All of the above�, etc., they all need to have a mark, so that the software knows that they have to be treated with a special logic. For example, if a choice was marked as �Does not apply�, this choice can be obviated in the calculation of the base for frequencies analysis. Another use of the mark is: if the choice �None of the above� was selected, the software will guarantee that the subjects who answered this choice cannot selected any other


3. Decide how you want to capture texts

Although Rotator offers the options: �short text� and �opinion text�, perhaps you may need to select the type �code to label�, with the purpose of saving time during data entry, because instead of typing the full answer, transcribers just need to type an equivalent code for it. For example, if a question was defined as �Opinion text�, a transcriber should type something like this: �I think the government has improved the education level in our country�, but if that question was defined as a �Code to label�, the transcriber should just need to type code �54�, which is equal to �Improved education� according to a pre-established table of codes.


4. Specify the �Analysis variables� of your survey

To tune your model, remember that those variables such as �Sex�, �Age�, Social stratum�, �City� and other social-psycho-demographic variables of single choice selection, must be defined as �Analysis variables�. In this way, Rotator will know that these variables should be placed on columns when you create a cross tab table, or for example, that they can be used to weighting your sample. (Among other benefits). To define a variable as �Analysis variable�, just edit the question, press the tab �Advanced� and click on the option �Is analysis variable�.


5. Specify the �Control variables� of your survey

For purposes of improving the administrative control of the survey, it is useful to identify control variables within your questionnaire such as �Name of Supervisor�, �Name of Interviewer�, �Interview Date�, �Street�, �Block�, etc. Namely, those variables that are not necessarily useful for analysis, but that are useful for controlling the survey process. To specify that a variable is a �Control variable�, just edit the variable, click on the tab �Advanced� and select the option �Is a control variable�.


6. Assign weights to choices (If applied)

If during the analysis you need to calculate an average, as well as other statistics from a single choice variable, it will be required that you assign a weight to each choice on that variable, because Rotator does not calculate statistics operations with choices codes (As SPSS does), but, with its weights. For example: If we have the variable �Age range�, with the choices: �Between 18 and 25�; �Between 26 and 30�; �Between 31 and 40� and �More than 40�, you can specify as a weight the midpoint of each interval: (Min + Max)/2. Thus for example, the interval �Between 26 and 30�, will have a weight of 28. Or for example, for the scalar variable �Satisfaction Level� with the choices: �Very satisfied�, �Satisfied�, �Somehow unsatisfied� and �Totally unsatisfied�, you may assign the weights: 2, 1, -1 and -2 respectively.


7. Check for errors
  • Verify that after compiling your survey there are no errors messages. If this occurs, please follow the recommendations displayed on the screen to correct those problems. In cases of extremely difficulty you may want contact the address shown at the end of this document. Remember that errors (RED) are severe and must be fixed urgently, warnings (YELLOW) must be fixed not so urgently, and recommendations (BLUE) might be obviated if you will.
  • If uncommon errors occur, such as: disfigurement of the questionnaire or PC hanging. These problems can be fixed by closing the application and opening it again (usually they are Windows problems. Use CTRL+ALT+DEL keys recommended by Windows to close the application).

8. Make Backups of your data and questionnaire
  • Please backup your model often. The possibility of fatal errors is always latent as with any other Microsoft application.
  • Organize with your MIS Administrator a daily backup of your Surveys folders.

9. Lock finished steps

On the lower side of each step there is an option called �Lock step� with the purpose of protecting the step against unexpected changes. To unlock a step just uncheck the step or go to the upper menu and select �Unlock all steps�. (Note: Only administrators can lock or unlock steps).


10. Share the survey folder on a LAN
  • If you want multiple transcribers or classifiers to simultaneously work at the same time with the same survey, just move your survey folder from your PC to a Local Network folder and share this folder with all the users involved. Once the folder has been shared, your survey can be accessed by all users created in the step �Team work�. To integrate data loaded by all users, just select the option �Consolidate� located at the step �Data�.
  • Coordinate with your Systems department to perform the necessary procedures to guarantee that the shared folders can be seen only by the users involved in your survey. And that the folders cannot be removed or deleted.

11. Test your questionnaire on the field and test some data
  • Aside from the methodology that you are using in your survey, we recommend you to test the questionnaire in the field before conducting the real survey. This test will allow you to detect if any question requires modification, if that question have to be deleted, or if is necessary to include instructions for the interviewers, etc. Additionally, you may measure the times that the interviewer takes; capture the reaction of the interviewees and interviewer, etc.
  • Once the questionnaire has been tested in the field, try to load some data to check the Data Entry process. If the process seems to be confusing or complex for the transcribers, try to assign an ALIAS code to each variable. Thus, instead of seeing the label �V35� you will see �Favorite Brand�. In this case, the alias codes must be short and mnemonic, because they will also be used for the construction of formulas later on.

12. Get some random data to check it on the Analyzer

Remember that you can generate random data to test your model in the �OLAP Surveys Analyzer� module. This is very useful since you will get a preview of how the data will be seen once the data entry process is finished. Just remember to delete that random data before starting the real load.