- On juillet 6, 2017
Here at Fusion Sport, we are always trying to help our clients optimise their data collection strategies. You may have read our previous blog post asking what are the optimal number Likert scale points?
But are Likert scales even the best type of survey instrument out there? Some of our clients have scrapped the use of Likert scales (and their 50 years of controversy) and opted for the ‘gold standard’ used in pain research: the Visual Analogue Scale (VAS).
The Problem with Likert Scales
Take this example of a fatigue Likert scale:
- Fully alert
- Very lively, but not at peak
- Okay, somewhat fresh
- Neither alert nor tired
- Somewhat tired, let down
- Very tired, very difficult to concentrate
- Completely exhausted, unable to function effectively
Does this scale represent a linear change in fatigue?
It has been long known that participants are less likely to choose the extreme options of a Likert scale. For many athletes, choosing option 7 over option 6 requires a greater difference in perceived fatigue than the difference in perceived fatigue between, say, options 4 and 3.
In essence, not all participants will interpret the same ‘distance’ between the Likert categories. Some academics will argue that it doesn’t matter: you can still assume that Likert data is roughly linear.
Nonetheless, read below to escape all the arguing statisticians and find out about a survey instrument that is already well-suited to wellness monitoring.
The Visual Analogue Scale
The Visual Analogue Scale (VAS) is pretty simple. All you need are two descriptive anchors at the two extremes (although there has not been any published research on VAS with multiple anchors).
Rather than ask participants to interpret the meaning of each Likert category, you could use a slider to represent a continuous relationship between the ‘anchors’. I like this approach because the instrument itself implies an underlying linear relationship.
In the context of wellness monitoring, I think this is an important trait since most wellness measures do appear intrinsically linear. Fatigue, soreness, appetite — i.e., common measures in many wellness surveys — would seem to be representing a continuum between general descriptors like ‘none at all’ to ‘completely maximised’. Indeed, some literature already recommends the VAS for the daily monitoring of fatigue and pain in clinical settings.
I would argue that Likert scales do not inherently imply linearity. There is certainly a hierarchy between the categories, but there is nothing in the construction of a Likert scale that says ‘by the way, options 1 and 2 are the same distance away from each other as options 5 and 6’. Some participants may see it that way, some may not.
The VAS is also more responsive. With more intervals, the VAS allows athletes to respond with greater precision. Likert scales on the other hand can often be an exercise in compromise; e.g. ‘I don’t quite fit in this category, or in this category, so which category do I choose?’ This will not be a problem with the VAS. The question is: compared to a Likert scale, does the increased precision of the VAS actually increase the resolution or does it just inflate the noise?
In 2012, Funke compared the response times of 413 participants’ use of Likert and VAS instruments and found that they were the same. Participants spent as much time answering questions using a VAS as they did using a Likert scale.
In saying that, the experimental group who used the VAS also made slightly more changes to their answers compared to the Likert scale users (measured by number of mouse clicks), implying that the VAS users were spending more time maximising the precision of their answers. To me these maximising efforts manifest in terms of data quality.
Some authors have found that VAS users tend to give slightly different responses compared to Likert users, but others have found no difference. What is certain is that the VAS has been shown to be equally reliable as Likert surveys again, and again, and again, etc.
All this suggests that, compared to Likert scales, the VAS does improve our ability to separate the signal from the noise — at least in daily monitoring settings.
In Smartabase, you could implement a VAS by using a slider question with 100 categories. That should be enough categories to imply a continuum.
In saying all that, there is much less research into the use of the VAS in elite sport contexts as there is research into Likert scales. Furthermore, best practice dictates that we should be training our athletes on the correct usage of survey instruments before they are implemented anyway. In other words, the linear vs. non-linear debate regarding Likert scales could potentially be mitigated by simply teaching athletes how they should approach daily monitoring instruments before using them.
If you are a Fusion Sport client who uses the Visual Analogue Scale in a wellness monitoring setting, we would love to hear from you. Do your athletes prefer using the VAS over Likert scales? Do you feel that the increased precision of the VAS improves data quality?