Author: Artee Gungah
Programming a survey entails more than merely replicating the paper version as an electronic format. There are various aspects you need to think about when programming the survey instrument. In this blog post, I am sharing seven tips to consider when programming electronic questionnaires. These tips are based on my experience of programming numerous large-scale, quantitative surveys over several years.
Questionnaire Layout – When programming the survey on your computer, remember the survey will be conducted by enumerators on a handheld device such as a tablet or phone. Therefore, you need to take aspects, such as the device’s size and the question’s structure, into account. Where possible, consider organising questions that are part of one module on one screen or page. This can help enumerators develop an overview of the modules and subsequently the questionnaire. As you programme the survey, it is advised you test it on the interviewing device to ensure it has been fully optimised and that enumerators can easily navigate and interact with the tool.
Wording of instructions – Instructions from the paper questionnaire used to inform enumerators to skip certain questions or perform calculations should be omitted altogether as these need to be part of the survey design. Likewise, you may need to edit wording in the electronic format, for example by asking enumerators to ‘enter’ rather than ‘write’ an integer response or to ‘tick’ rather than ‘circle’ a multiple-type response.
Formatting – You can customise the format of your question texts and instructions by using HTML, a simple programming language. For instance, to emphasize specific words, instructions, or question texts you can show the texts in bold. You should also consider displaying enumerator instructions in a specific colour so that enumerators are alerted straight away as to which texts should not be read out to respondents. Whichever format you decide to opt for, always keep them consistent throughout your questionnaire.
Validation checks – Before writing validation checks, consult with a local expert who is knowledgeable on the subject matter of the survey as they are more likely to know about possible answers that can be entered. Validation checks can be based on answers given or calculations made during a live interview and on data collected during previous surveys. When writing validation checks, you need to be especially careful if the CAPI software you are using will not let enumerators move forward with the interview if a validation error is found. In such cases, it is important to account for unusual but possible answers when writing validation checks. Finally, messages linked to validation checks should be short and clear; they should be able to promptly guide the enumerator on which answer needs to be corrected or confirmed by referring to the exact question.
Hidden fields – You can add extra variables in your questionnaire to store values or automatically perform calculations instead of having enumerators do them. Some of these hidden fields are only used to make the design of the survey easier and they often need to be dropped from the final data set. When naming these variables, consider using a prefix such as ‘hidden_’ to flag them. They can later be grouped and easily dropped from the final data set using statistical software.
Identifiers at different levels of observation – Typically, the CAPI software will automatically generate IDs to uniquely identify data at different levels of observation within your exported data set. However, these are often not easily readable. When programming the survey, consider adding extra variables to not only store the identifiers at the current level of observation but also any parent identifiers. For example, if ‘regionID’ and ‘householdID’ are part of your study IDs at the highest level of observation and you add a roster to record details about household members, consider adding extra variables to store the region and household identifiers at the roster level as well. This way, if you export your roster data in a separate data file, you will find this information there as well which will help you easily and uniquely identify your data.
Test your electronic survey – Thoroughly test your electronic survey to make sure the tool is as error-free as possible. The testing process becomes less cumbersome if you test the survey as you programme it.
If you found these tips useful, you can also check out our blogs on Merging Data Sets at Different Levels in Stata and Tips for Conducting COVID-safe Focus Group Discussions.