How to build an online questionnaire
Building an online questionnaire
– everything you need.
This tutorial contains everything you need to learn how to build and run your own online surveys and questionnaires. As a bonus you’ll also see how choice experiments work.
(this guide follows the video above)
This guide will show you how to setup and collect questionaire data using the SurveyEngine software. You will also learn how to include a simple choice experiment.
Knowing the real-world issue and limitations and their mitigations will affect how you plan your study.
You should have already registered or logged in and be looking at the dashboard screen, everything we’ll do today can be achieved with the free version.
The dashboard shows you all your current projects and their status. From here you can dive in and manage a project. You can or create a new project or import one you’ve saved offline or downloaded from the tutorials.
You can also import and export entire surveys as Excel spreadsheets. Which is useful if you want to create surveys offline in excel from data bases such as like a products or medications data rather than entering them manually.
Building the Survey
Firstly create a blank survey with the ‘create survey’ button on the top right.
Note, within each survey project, there are three phases: the editing phase, where you’re building a survey, entering questions and text and getting the logic and flow right. Then there’s the data collection phase, where you’r monitoring respondents coming in. Finally there’s the results – running analysis and downloading data.
The first screen you see after you’ve created the survey is the page editor and this is the screen that you’ll probably spend most of your time on. On this screen you can create pages, place questions, text, images. logic and choice scenarios texts and logic within those pages.
You start with blank pages and you can always add more or delete them. For the purposes of this example we’ll imagine we’re running a survey on route choice, and we’ll also put a small choice experiment in. We’ll also collect some demographic and other data to control for hypothetical bias.
So we’ll start by putting some text and introduction on our first page welcoming the respondent to the survey, giving them an indication of the subject matter and how long the survey will take. On the first page we’ll also add a consent question and some logic branching them to a screen out page if they don’t consent.
Data Labels and Expressions
All questions have data labels. This gives the data collected a meaningful name which will appear in the exported data. It will also allow us to use it in the survey. In this case we’ll use it as a condition to branch the user to another page. You can always see all the variables used in the survey in the Dictionary screen.
The language used is called Perl and for which there are many online resources to learn. You can include whole Perl programs but today we’re just using it to test and branch.
Adding a branching element that directs the flow to the final page. Without any conditional logic, it will direct all users to the final page, but by adding a condition to, we can control who we send to the screenout page. Specifically we want the branching to be enabled if they user selects the second option for consent: i.e. ‘No’.
Now on the second page we’ll add a question about car ownership, which we may use in our models to segment later. Placing 4 place holders for the experiment which we’ll do later and a final thank you and comments question.
At this point, you might like to preview the survey to see the logic is working.
A Simple Choice Experiment
TO create a simple route choice experiment, we use the experiment screen and then come back, adding the experiment scenarios just like regular questions.
In the experiment editor entering 2 alternatives and then 3 attributes we are interested in: travel time, toll costs and congestion.
- travel time: 20min, 40 mins, 1 hour.
- congestion: no congestion, infrequent congestion, frequent
- toll cost: none, $2.00, $5.00 and $10.00
Finally generating an experiment design you should review the layout and you’re done.
Back in the survey builder just replace the placeholders with scenarios. Typically you would add some text and data to help control for hypothetical bias.
For example, the following text would go some way to promoting incentive compatibility.
“Results of this study will be used to develop roads and tolls in your city. It is important that you make choices as you would make in a realistic setting”
Also a question at the end may help control for hypothetical bias in analysis.
“How certain were you of your choices”: “Very certain, certain, uncertain, very uncertain and don’t know.”
Final adjustments to the layout and previewing to confirm it looks fine and you’re ready for the deployment screens.
Publishing a Survey
The publish screen is the gateway screen to getting your current survey live. When you publish, a copy is made of your current working edits. This is put live with a version number starting at one. This is important for management later on If you, for example, make small changes to your survey during data collection but want the data to be merged but to be still able to identify responses by version.
Or if you’re running a series of tests and you want to clear the data out and start fresh. Every new publish is a new version. And the version history screen can show you all your published versions and allow you to include o exclude data from those version.
Clicking publish makes this version live and takes you to the Activity Screen.
Live Activity Screen
On this screen you can see the summary of completion statistics for the survey. The first time you do this all activity and completion stats will be zero. As respondents come in you’ll be able to see who’s live, who’s been screened out and the total completes. With a more complex stratification with quotas for different segments each will be displayed here as well.
To get your survey live, and for that you will need an invitation link that you use in either
- an outbound email campaign or
- as a post on your site or social media, or
- given to a panel provider
You can create multiple links which is usefuel If you want to track and manage those multiple multiple sources. These invitation links allow you to dictate where people go after they finish the survey which is critical when dealing with panels who need to redirect back to their site to count completions and screenouts.
Ideally you want to test you survey before going live. Doing this manually is burdensome so there is an automated simulation screen that generates bots to test your whole survey before going to field.
These bots run through your survey just like a respondent; ccanning each page the page looking for active buttons and fields which they click and enter. From a data collection perspective. They are identical to real respondents. With the only difference is that their IP addresses are all the same, and they gave random answers.
But in all other ways they will actually be interacting with the survey as a respondent woudl. Any logic errors or even any technical errors with code will be experienced by them just like a real respondent.
They also will generate data so you can see what the data will look like, which is useful to find any holes and even to prepare your modelling scripts.
Once you’ve collected your data, you’llcan examine it download it. This can be done at any time during the data collection. Downloading the data is a matter of going to the data screens and downloading it in the format you want, SPSS, CSV or Excel. Each download also includes data dictionaries.
Analysis and Reports
There’s a frequency and a cross tab screen that lets you do a quick analysis without having to download the data. A more permanent version of these analyses can be performed as a snapshot and shared as online HTML pages.
The modelling screen lets you generate a pre-defined MNL model for each choice experiment in your survey. While its recommended to download the data and perform your own modelling, this screen gives you an early read on model convergence.
For more detail on using the SurveyEngine platform, visit the documentation pages.