2

I'm debating whether to use a radio button or a slider for a survey. My thoughts at the moment are the following. I'd like the group's views and / or links to research which has studied something similar.

The question will ask for agreement on a statement. Here are my thoughts so far:

Radio buttons (5 point scale)

Advantages

  • They're conventional / well known
  • They are easy to use on touch and non-touch devices

Disadvantages

  • We're forcing the user to 'bin' their perception, which is on a continuum, to one of 5 positions
  • The data needs to be treated as ordinal & as there are only 5 possibilities the available analytic / visualisation choices are limited
  • labelling. Is 'strongly agree' 2X stronger emotion than 'agree' or 1.5X (or any other amount). We know there are cultural differences in Likert selection. Multiple languages makes this even more of an issue.

Slider (Continuous, two extremes labelled, no labelling on position)

Advantages

  • The control is a continuum, which reflects the users perceptions
  • I think I can more safely assume that the user interprets the position as linear and can therefore treat the values as interval

Disadvatages

  • There is an accuracy issue with the data (are two questions slightly with slightly different positions meant to be the same?) This is likely to depend on the device.
  • There is less guidance / feedback in the form of labelling for the user.
  • Depends on the question. If you're asking for an opinion, a slider that would go from "Bad" to "Good" (for example) would be weird. What would be the difference between one slider position and another one a few pixels away? If you're asking for some numerical data, a slider could work for a question like "How long does it take you to go to work?". In any case, a slider will provide you with a lot more data, that will take more time to parse. – jgthms Feb 26 '14 at 12:19
  • I've done a bit more reading on this. What I've created with the slider is a 'semantic differential' scale. Historically the most common number of points on this scale seems to be 7. I don't know whether this is because that's best for the respondent or the analyst. As the analyst more data outweighs effort in analysis, as long as the data is reliable / issues can be dealt with – AndrewMinCH Feb 26 '14 at 13:20
0

I'd make sure it worked with your question, for instance if you were asking someone how happy they were and you had a slider setup like this

:( --------------------------[[]]---------------------------- :)

You would undoubtedly get a scattered data set back but it would still be usable with bar charts or other methods.

One thing to note about sliders is that on mobile you need to optimize for the human finger aka make your touch targets large enough to easily be used in android the minimum is 48dp. As well make sure that your slider is not laggy because that will at-least turn people like myself away.

Hope this helps!

  • 1
    Thanks for this. I did quite a bit of research on this topic and found this paper which had studied an almost identical problem: iscience.deusto.es/wp-content/uploads/2011/09/… It seems that the key point to increase usability of sliders is being to adjust them by selecting (e.g. tapping) rather than just by sliding. As you note the data is scattered and there is a measurement error, but there is a measurement error with existing radio-buttons anyway. We'll do some analysis on estimating this depending on the device type. – AndrewMinCH Mar 14 '14 at 8:34

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.