Anyone know what kind of data is this scale ? interval, ratio or ordinal? I really have no idea!

#1
Hi!, Does anyone come across a scale like the one below, which involve with graph and is different from usual scale. May I know if anyone know what kind of data is thisscale ? interval, ratio or ordinal?

12571_2020_1115_Fig1_HTML.png
 

Karabiner

TS Contributor
#2
The categories are ordered according to degree of overlap, i.e. ordinal scale.

With kind regards

Karabiner
 
Last edited:
#3
The categories are ordered accoring to degree of overlap, i.e. ordinal scale.

With kind regards

Karabiner
Oh, thanks ! but the thing is I need to combine these two scales into one variable, as you can see in the graph, one is ordinal as you said, and the other is Likert scale, what should I do if I want to combine them? 1122.jpg
 

Karabiner

TS Contributor
#4
Does the second scale consist of three items which are answered on a rating scale with 7 levels?
What does that second scale actually measure?

And why do you want to combine the two measuerements - what is idea behind that, and what do
you want to do with the resulting combination?

With kind regards

Karabiner
 
#5
Does the second scale consist of three items which are answered on a rating scale with 7 levels?
What does that second scale actually measure?

And why do you want to combine the two measuerements - what is idea behind that, and what do
you want to do with the resulting combination?

With kind regards

Karabiner
Yes, as you can see, the second scale consists of three items measured by 7 levels with likert scale. The reason that I want to combine two measures is because I these two measures are the two main components of the social identity variable. Eventually, I want to measure the relationship between media use and society identity.
 

Karabiner

TS Contributor
#6
What is the correlation coefficient between the 3-item-scale and the item with the graphical display?
And could you maybe describe how the items of the 3-item-scale are formulated?

With kind regards

Karabiner
 
#7
What is the correlation coefficient between the 3-item-scale and the item with the graphical display?
And could you maybe describe how the items of the 3-item-scale are formulated?

With kind regards

Karabiner
I attached the graphical display here.. As you can that, I use two scale to measure the latent variable, social identity, including Inclusion of community in self and social identification based on pervious literature. Therefore, I need to combine the results of these two scales, ang get the mean of the combined variable, which manifests social identity.

conceptual model-Page-2.jpg
 

Karabiner

TS Contributor
#8
It would be interesting to know whether inclusion & identification are highly, moderately, or scantily correlated.

You could prevent the pains of collapsing both aspects into one measure by using a multivariate approach, i.e.
you use both measures as dependent variables in a multivariate analysis of variance (MANOVA), with social media
use
as predictor (in SPSS MANOVA, it would be called either "factor" or "covariate", depending on the scale level).
This would presuppose that you can use the inclusion variable as interval scale.

With kind regards

Karabiner
 
#9
It would be interesting to know whether inclusion & identification are highly, moderately, or scantily correlated.

You could prevent the pains of collapsing both aspects into one measure by using a multivariate approach, i.e.
you use both measures as dependent variables in a multivariate analysis of variance (MANOVA), with social media
use
as predictor (in SPSS MANOVA, it would be called either "factor" or "covariate", depending on the scale level).
This would presuppose that you can use the inclusion variable as interval scale.

With kind regards

Karabiner
I see, many thanks!. but if I want to combine these two variable, all I can do is to transform interval data( social identification) to ordinal(inclusion in self) , not the other way around, right? given that I cannot transform lower level data to higher, right?
 

noetsi

No cake for spunky
#11
You can combine likert data in a metric, for example factor analysis. At a certain point it is treated as interval even if formally it does not meet those definitions. When you have a 15 unique level likert scale it is going to act like interval data. When you combine it more and get a hundred point scale it is going to act even more interval like :)
 
#12
It would be interesting to know whether inclusion & identification are highly, moderately, or scantily correlated.

You could prevent the pains of collapsing both aspects into one measure by using a multivariate approach, i.e.
you use both measures as dependent variables in a multivariate analysis of variance (MANOVA), with social media
use
as predictor (in SPSS MANOVA, it would be called either "factor" or "covariate", depending on the scale level).
This would presuppose that you can use the inclusion variable as interval scale.

With kind regards

Karabiner
May I ask you one more question. In regression analysis, if I need to know the total effect of a latent construct, people normally assume that different measured variable in a latent variable are equally important, so they calculate a weighted average and then summed them to create a weighted construct for their latent variable. However, in structural equation model(SEM), I wonder if SEM also treat different measured variables that indicate the latent variables equal too.