Likert Scale Analysis

Discussion:

(too old to reply)

Julia

2005-05-17 00:21:39 UTC

I am using a 5-point Likert Scale to measure students' experiences at
school (ex. relationships with teachers, fear of academic failure,
etc.) to determine whether students with poor attendance have more
negative experiences at school than students with regular attendance. A
sample item is "I feel like I don't belong at school." I am
assuming that students with poor attendance will have higher mean
scores than students with regular attendance.

What is the best way to analyze this data? I am going to treat the data
as interval data.

Any advice is gratefully appreciated.

Julia

'`'`'`'`'`'`'`'`'`'`'`'`'`''`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`
sci.psychology.research is a moderated newsgroup.
Before submitting an article, please read the guidelines which are posted
here bimonthly or the charter on the web at http://psychcentral.com/spr/
Submissions are acknowledged automatically.

s***@yahoo.com.au

2005-05-22 02:19:06 UTC

Permalink

Julia, the best way to analyse the data is to not treat them as
interval. The "numbers" in a 5-point Likert "scale" are really just
categories. Hopefully, they are ordered, in which case the data are
ordinal. They are not, however, interval.

Best way to start analysing the response data is by applying the Rasch
measurement model. They can be analysed easily in software such as
RUMM2020 (www.rummlab.com.au) or another other Rasch software package.
By applying the Rasch model, you can estimate the locations of students
on the scale. A higher location would indicate more positive
experiences.

If you are using a "neutral" category, be warned that this category
often fails to function properly, in the sense that it does not
effectively capture an attitude to a statement that is between
"disagree" and "agree". You're better off using a four point scale,
excluding the middle category.

It is a scientific hypothesis that higher category responses reflect
more positive experiences, attitudes, or whatever is intended to be
captured in the statements. This hypothesis can be tested by applying
the Rasch model. There's no justificaiton for treating Likert scales
as interval measurements: don't let the widespread nature of such
practices fool you.

With locations estimated, you can apply statistical tests or models.
Then it depends on what sort of attendance data you have, and what
other variables you have.

Personally, I would not use Likert scales for a start. They generally
fail to deal with context, in the sense that different people can
imagine different contexts when they are reading the statements. They
also lend themselves to much to response set, which often results in
apparent but false reliability. Nonetheless, if you've done so,
applying the Rasch model will still tell you a great deal.

Take care,
Steve

Post by Julia
I am using a 5-point Likert Scale to measure students' experiences at
school (ex. relationships with teachers, fear of academic failure,
etc.) to determine whether students with poor attendance have more
negative experiences at school than students with regular attendance. A
sample item is "I feel like I don't belong at school." I am
assuming that students with poor attendance will have higher mean
scores than students with regular attendance.
What is the best way to analyze this data? I am going to treat the data
as interval data.
Any advice is gratefully appreciated.
Julia

'`'`'`'`'`'`'`'`'`'`'`'`'`''`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`

Post by Julia
sci.psychology.research is a moderated newsgroup.
Before submitting an article, please read the guidelines which are posted
here bimonthly or the charter on the web at

http://psychcentral.com/spr/

Post by Julia
Submissions are acknowledged automatically.

'`'`'`'`'`'`'`'`'`'`'`'`'`''`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`
sci.psychology.research is a moderated newsgroup.
Before submitting an article, please read the guidelines which are posted
here bimonthly or the charter on the web at http://psychcentral.com/spr/
Submissions are acknowledged automatically.

Tamarah

2005-05-27 01:53:01 UTC

Permalink

Steve,

I see you know something about RUMM software and wondering if you can
help me. I am having difficulty using it for the first time. Can you
tell me how to enter the data for the individual scores/responses. I
have the blocks and items all set up. The program goes right into the
analysis stage, but there's no data!? Can you help?
Thanks,
Tamarah

Post by s***@yahoo.com.au
Julia, the best way to analyse the data is to not treat them as
interval. The "numbers" in a 5-point Likert "scale" are really just
categories. Hopefully, they are ordered, in which case the data are
ordinal. They are not, however, interval.
Best way to start analysing the response data is by applying the Rasch
measurement model. They can be analysed easily in software such as
RUMM2020 (www.rummlab.com.au) or another other Rasch software package.
By applying the Rasch model, you can estimate the locations of students
on the scale. A higher location would indicate more positive
experiences.
If you are using a "neutral" category, be warned that this category
often fails to function properly, in the sense that it does not
effectively capture an attitude to a statement that is between
"disagree" and "agree". You're better off using a four point scale,
excluding the middle category.
It is a scientific hypothesis that higher category responses reflect
more positive experiences, attitudes, or whatever is intended to be
captured in the statements. This hypothesis can be tested by applying
the Rasch model. There's no justificaiton for treating Likert scales
as interval measurements: don't let the widespread nature of such
practices fool you.
With locations estimated, you can apply statistical tests or models.
Then it depends on what sort of attendance data you have, and what
other variables you have.
Personally, I would not use Likert scales for a start. They generally
fail to deal with context, in the sense that different people can
imagine different contexts when they are reading the statements. They
also lend themselves to much to response set, which often results in
apparent but false reliability. Nonetheless, if you've done so,
applying the Rasch model will still tell you a great deal.
Take care,
Steve

Post by Julia
sample item is "I feel like I don't belong at school." I am
assuming that students with poor attendance will have higher mean
scores than students with regular attendance.
What is the best way to analyze this data? I am going to treat the

data

Post by Julia
as interval data.
Any advice is gratefully appreciated.
Julia

'`'`'`'`'`'`'`'`'`'`'`'`'`''`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`

Post by Julia
sci.psychology.research is a moderated newsgroup.
Before submitting an article, please read the guidelines which are

posted

Post by Julia
here bimonthly or the charter on the web at

http://psychcentral.com/spr/

Post by Julia
Submissions are acknowledged automatically.

s***@yahoo.com.au

2005-05-27 11:07:42 UTC

Permalink

Post by Tamarah
Steve,
I see you know something about RUMM software and wondering if you can
help me. I am having difficulty using it for the first time. Can you
tell me how to enter the data for the individual scores/responses. I
have the blocks and items all set up. The program goes right into the
analysis stage, but there's no data!? Can you help?
Thanks,
Tamarah

Tamarah, you need to select the data file early in the process, right
after you have created a new project and given it a name. Once you
have slected "new project", and named it, RUMM will prompt to you for a
project name. Just leave the default name here, and click on apply.
Next, it will bring up a window for opening a data file, in which you
can browse. At this point, you need to select your data file. Then
you can proceed to data design, data format, etc.

If you save the data file from Excel, save it as a .prn file with one
column per item. It needs to be a text file that is not tab-delimited.

You say you have the blocks and items all set up - do you mean the
template files (i.e. the .itm for person background info and blocks and
the .spc for type of item, number of categories etc.)?

Steve

'`'`'`'`'`'`'`'`'`'`'`'`'`''`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`
sci.psychology.research is a moderated newsgroup.
Before submitting an article, please read the guidelines which are posted
here bimonthly or the charter on the web at http://psychcentral.com/spr/
Submissions are acknowledged automatically.