Analysis for variables on different likert scales

[Kuan]

Hi,

I measured two different variables measured on two different scales: one is
on a 7 point likert scale (1 thru 7) and the other is on a 4 point likert
scale (1 thru 4). Is it adequate to run regression analysis on the original
scores of the two variables? Do I need to use their standardized scores for
running the analysis? Any suggestions would be highly appreciated.

Kuan

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Ryan]

Kuan,

You haven't given us much to work with, except that you have two
variables that are measured on different scales, and that you're
interested in running a "regression analysis." I'm guessing you want
to estimate the linear relationship between two variables. In general,
you do not *need* to standardize variables before entering them into a
linear regression analysis, even if they are measured on different
scales. By standardizing one or both of the variables, interpretation
of the beta coefficients will change, but the correlation coefficient
will remain the same.

Ryan

On Mon, Jul 11, 2011 at 11:54 AM, Kuan <[hidden email]> wrote:

> Hi,
>
> I measured two different variables measured on two different scales: one is
> on a 7 point likert scale (1 thru 7) and the other is on a 4 point likert
> scale (1 thru 4). Is it adequate to run regression analysis on the original
> scores of the two variables? Do I need to use their standardized scores for
> running the analysis? Any suggestions would be highly appreciated.
>
> Kuan
>
> =====================
> To manage your subscription to SPSSX-L, send a message to
> [hidden email] (not to SPSSX-L), with no body text except the
> command. To leave the list, send the command
> SIGNOFF SPSSX-L
> For a list of commands to manage subscriptions, send the command
> INFO REFCARD
>

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Kuan]

Hi Ryan, Thanks for your suggestion.

Actually, I would like to develop a model with a small sample size. I have
several variables measured on 7 point likert scale and only one variable
measured on 4 point likert scale. Because of a small sample size, I consider
running multiple regressions to create the model. Hence, is it more
appropriate to standardize all the variables for developing the model?

Kuan

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Rich Ulrich]

I've used variables standardized to mean=50, SD=10  as a
convenient way to show the range of composite scores.
That is for *display*, and not for their effective *use*  in
analyses.

About the only *use* I have made of standardizing variables is
when I give nominally-equal weights to the parts when I create
a composite score.  That is straightforward.  I do not see
where you have any incentive to use standardized scores as
part of "running multiple regressions to create [a] model."

What are you thinking of, that makes that seem appropriate?

--
Rich Ulrich [

> Date: Tue, 12 Jul 2011 06:31:20 -0400

> From: [hidden email]
> Subject: Re: Analysis for variables on different likert scales
> To: [hidden email]
>
> Hi Ryan, Thanks for your suggestion.
>
> Actually, I would like to develop a model with a small sample size. I have
> several variables measured on 7 point likert scale and only one variable
> measured on 4 point likert scale. Because of a small sample size, I consider
> running multiple regressions to create the model. Hence, is it more
> appropriate to standardize all the variables for developing the model?
>
> Kuan

[snip, sig]

[Ryan]

Kuan,

Standardizing variables before entering them into a linear regression
will not alleviate problems related to small sample sizes. I have a
feeling I'm misunderstanding you.

Ryan

On Tue, Jul 12, 2011 at 6:31 AM, Kuan <[hidden email]> wrote:

> Hi Ryan, Thanks for your suggestion.
>
> Actually, I would like to develop a model with a small sample size. I have
> several variables measured on 7 point likert scale and only one variable
> measured on 4 point likert scale. Because of a small sample size, I consider
> running multiple regressions to create the model. Hence, is it more
> appropriate to standardize all the variables for developing the model?
>
> Kuan
>
> =====================
> To manage your subscription to SPSSX-L, send a message to
> [hidden email] (not to SPSSX-L), with no body text except the
> command. To leave the list, send the command
> SIGNOFF SPSSX-L
> For a list of commands to manage subscriptions, send the command
> INFO REFCARD
>

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD