Likert scales

#1
Hi. I have a couple of questions about the controversial Likert scales.
1) what are the "acceptable" statistical analysis procedures for the responses of one sample. No comparison of groups.
2) can the same scale have both agreement and frequency items? Or the anchor must be fixed eg. totally disagree to totally agree.
3) When exploratory factor analysis is applied, what is the best extraction and the best rotation methods you suggest?
Thanks in advance for your contributions...
 
Last edited:

Karabiner

TS Contributor
#2
Hi. I have a couple of questions about the controversial Likert scales.
Do you really mean Likert scales (consisting of several Likert-type items), or just a single Likert-type items, with a Likert-type response format? I am assking this because these two are often confused.

1) what are the "acceptable" statistical analysis procedures for the responses of one sample. No comparison of groups.
This depends on the research question. Which one(s) do you have in mind?

2) can the same scale have both agreement and frequency items? Or the anchor must be fixed eg. totally disagree to totally agree.
I am not quite sure what you are after. Have you read definitions of Likert scales already?

When exploratory factor analysis is applied, what is the best extraction and the best rotation methods you suggest?
In which context?

Thanks in advance for your contributions.
Are these homework questions? What are your own ideas, then?

With kind regards

Karabiner
 
#3
Such a typical response of some guys who think they know the answers to all existential riddles. I would not have asked such questions if I did not know the difference between the items and the scale. My second question has nothing to do with the definition, and I believe it is clear to anyone who has some background in English and statistics. And for the philosophical comment about context, EFA does look for the context; it does not depend on it. It seems that you are the one who should do some homework before you decide to comment with some useless nonsense. Save your time and mine, please.

All the best
 
Last edited:

ondansetron

TS Contributor
#4
Such a typical response of some guys who think they know the answers to all existential riddles. I would not have asked such questions if I did not know the difference between the items and the scale.
@Karabiner asked fair questions. Typically people conflate Likert items and scales and most people who think they're using these are actually using Likert-type items/scales because they don't meet the theoretical requirements of true Likert items/scales. Getting defensive about your knowledge isn't a good place to start when someone is trying to make sure they understand your question.

My second question has nothing to do with the definition, and I believe it is clear to anyone who has some background in English and statistics.
Again, you can knock the guy for trying to clarify your question, or you can provide more information. We're only trying to help. I agree your statement is unclear; I'm a native English speaker and have "some background" in statistics (although, we are unclear what that means).

And for the philosophical comment about context, EFA does look for the context; it does not depend on it.
To state that any statistical analysis doesn't depend on context is a long-shot, so let's not make that claim. EFA, like CFA, is often a mystical, hand-wavy analysis in the majority of cases. Few are truly performed well. EFA will not provide you with context, it provides you with some quantification of a covariance structure in a particular set of variables--you need to give context and meaning to the "factors" that are supposedly driving the covariance structure.

It seems that you are the one who should do some homework before you decide to comment with some useless nonsense. Save your time and mine, please.

All the best
I've always gotten the most help when I react defensively to people asking for clarification.
 
Last edited:

Karabiner

TS Contributor
#5
Such a typical response of some guys who think they know the answers to all existential riddles.
I have a different view.

I would not have asked such questions if I did not know the difference between the items and the scale.
That is something that you personally knew about yourself, but it wasn't something
which someone here could be sure of. As I wrote before, Likert scale and Likert items
are often confused by those asking for help here. You knew the difference, so it's fine.

My second question has nothing to do with the definition, and I believe it is clear to anyone who has some background in English and statistics.
So if you just could be so kind as to explain what your question actually meant?

And for the philosophical comment about context, EFA does look for the context; it does not depend on it.
The question concerned the practical and/or theoretical problems which EFA should
solve. Which of the several extraction and rotation methods is "best" depends on the
objective of the analysis and on the data at hand.

It seems that you are the one who should do some homework before you decide to comment with some useless nonsense. Save your time and mine, please.
That does not answer the question whether this is homework. Or what you
suggest as solutions.

With kind regards

Karabiner
 
#6
TO: ondansetron
Nothing is defensive about my reply; it is just descriptive.
One of the problems of such forums is that you never know the real identity of members and how much they really know about what they think they know.
Agreement response items and frequency response items are usually used in Likert. And the question is whether the same scale can have a mix of both or it must use one response category. Hope it is clear enough now. Maybe the word "anchor" is the problem but it is used normally in the literature.
EFA does help with finding the trends among correlations, which constitute factors that are the pillars of that context. It explores what is there..it does not rely on it. I am sure you know there are many cases of statistical analysis that do not rely on the context at all.
Once again I described. I did not defend

TO:Karabiner
Let's say I am asking 400 participants about their attitudes to life in the city. No homework here.
 

ondansetron

TS Contributor
#7
TO: ondansetron
Nothing is defensive about my reply; it is just descriptive.
One of the problems of such forums is that you never know the real identity of members and how much they really know about what they think they know.
TO: DarioMarenelli This is why we ask questions...


EFA does help with finding the trends among correlations, which constitute factors that are the pillars of that context. It explores what is there..it does not rely on it. I am sure you know there are many cases of statistical analysis that do not rely on the context at all.
Once again I described. I did not defend
Finding them on EFA doesn't make them real, they very well can be spurious covariance structures, so context is needed, even in exploratory. Give some examples of statistical analyses that don't need context.

You also used ad hominems on Karabiner, so I'd say you were defensive.