Statisticians - one...
 

  You don't need to be an 'investor' to invest in Singletrack: 6 days left: 95% of target - Find out more

[Closed] Statisticians - one way Likert scales

7 Posts
4 Users
0 Reactions
89 Views
Posts: 7812
Full Member
Topic starter
 

Just interested whether thinking has changed on how to use Likert scales since I was an undergraduate.

We were always encouraged to "reverse" some of the questions but to be a bit careful about how..

E.g.

I think the NHS should be privatised
Vs.
Healthcare should be free to all

Then you reverse scored them to get the answers the same way round to understand what's behind them.

So I would expect me to say strongly disagree on first one and strongly agree on second.

The idea was you didn't have a "right" side and "wrong" side of the scale for the replies for each person, which was supposed to help with validity or some such.

I'm seeing a lot of one sided stuff lately in surveys etc. Just wondering if the world of research / stats has moved on or is it just a dying art/the malaise of the modern world?

Sorry totally nerdy question. 🙂


 
Posted : 19/07/2021 11:37 pm
Posts: 0
Free Member
 

The surveys are no longer objective.


 
Posted : 19/07/2021 11:43 pm
Posts: 7812
Full Member
Topic starter
 

The surveys are no longer objective.

As in they don't work at a statistical level by going one sided or they're being deliberately used in a different way?

I was trying to read around this and my feeling is that if the completer/subject knows the "desired" outcome you would be more likely to get some unconscious acquiescence if it was one sided to the desired answer.

I know we used some for a survey at a sports club I'm involved in a while back.


 
Posted : 19/07/2021 11:53 pm
Posts: 11961
Full Member
 

Ultimately, you need to run a psychometric analysis to see if it's functioning as intended. Having reversed items can help with that, but you need to actually pilot it and check that the items are functioning properly.

Another source of problems is the number of categories on the scale. Researchers often think that including more categories increases the precision, but at some point, it becomes counterproductive. Four or six categories often works better than 10 because respondents are more likely to use the full range of the scale when you have fewer categories. Also, having an odd number of categories encourages people to just choose the middle category, whereas an even number forces a choice. Ultimately, you need to actually conduct a psychometric analysis to see what's happening because people often don't respond the way you assume they would.

Also, to be pedantic, statistics and psychometrics aren't exactly the same thing. Statisticians compare sets of numbers. Psychometricians measure psychological properties.


 
Posted : 20/07/2021 12:08 am
Posts: 8904
Free Member
 

We were always encouraged to “reverse” some of the questions but to be a bit careful about how..

E.g.

I think the NHS should be privatised
Vs.
Healthcare should be free to all

Are those actually reversed questions though?
Possible to agree with both 'I think private companies would run it better than the state and this doesn't matter if its still free at the point of use' or to disagree with both 'it should be state run and not for profit but elective procedures such as IVF should be charged for'


 
Posted : 20/07/2021 12:32 am
Posts: 0
Free Member
 

they’re being deliberately used in a different way?

Yes

I was trying to read around this and my feeling is that if the completer/subject knows the “desired” outcome you would be more likely to get some unconscious acquiescence if it was one sided to the desired answer.

I reckon. Not sure how much of it is unconscious though.
The polls/surveys/votes/whatevers are also targeted, so you're likely to see huge numbers in favour of given outcomes. Seems to be a trump/post trump thing IMO. Good practice/statistical significance/truth doesn't matter, it's about having data to back up what YOU want to be true. There are good surveys out there - there is just a lot more noise/false positives.

Nothing to back this up, just what I've noticed in the last few years. Do a lot of stats work, but I'm a chemist, so nothing as wooly as surveys etc. I might be well off the mark.


 
Posted : 20/07/2021 12:34 am
Posts: 8904
Free Member
 

https://m.youtube.com/watch?v=G0ZZJXw4MTA


 
Posted : 20/07/2021 12:37 am
Posts: 7812
Full Member
Topic starter
 

Thanks all. Curiosity satisfied and some interesting points and don't worry andrewh I'm not writing one about the NHS 🙂


 
Posted : 20/07/2021 7:26 am

6 DAYS LEFT
We are currently at 95% of our target!