SPSSX Discussion

Testing for Sig. Dif. between 2 Ordinal Variables

Classic

List

Threaded

8 messages Options

Rich Goldman-3

Testing for Sig. Dif. between 2 Ordinal Variables

Hi. I have survey data in which several variables represent the reported
frequency with which participants do or use something. For example:

How regularly do you use the telephone for communication with your peers?
Never (1)
Rarely (2)
Sometimes (3)
Often (4)
Very Frequently (5)

I want to measure if there is a significant difference in the frequencies
reported between 2 (or more) such variables (e.g. is there a sig. dif in the
frequency the telephone is used vs. email) and ideally the direction of the
difference. The variables being compared use the same never-very frequently
1-5 ranking.

What tests would people on this list recommend using to do this? Thanks for
any help you can provide.
-Rich

=====================
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

Swank, Paul R

Re: Testing for Sig. Dif. between 2 Ordinal Variables

There is the Stuart extension to the McNemar test which tests for equality of correlated marginal probabilities in a square contingency table. Google it.

Dr. Paul R. Swank,
Professor and Director of Research
Children's Learning Institute
University of Texas Health Science Center-Houston

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Rich Goldman
Sent: Tuesday, March 03, 2009 1:05 AM
To: [hidden email]
Subject: Testing for Sig. Dif. between 2 Ordinal Variables

Hi. I have survey data in which several variables represent the reported
frequency with which participants do or use something. For example:

How regularly do you use the telephone for communication with your peers?
Never (1)
Rarely (2)
Sometimes (3)
Often (4)
Very Frequently (5)

I want to measure if there is a significant difference in the frequencies
reported between 2 (or more) such variables (e.g. is there a sig. dif in the
frequency the telephone is used vs. email) and ideally the direction of the
difference. The variables being compared use the same never-very frequently
1-5 ranking.

What tests would people on this list recommend using to do this? Thanks for
any help you can provide.
-Rich

=====================
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

David Greenberg

Re: Testing for Sig. Dif. between 2 Ordinal Variables

In reply to this post by Rich Goldman-3

A chi-square test will do this, but it won't give you information about direction. David Greenberg, Sociology Department, New York University

----- Original Message -----
From: Rich Goldman <[hidden email]>
Date: Tuesday, March 3, 2009 12:13 pm
Subject: Testing for Sig. Dif. between 2 Ordinal Variables
To: [hidden email]

> Hi. I have survey data in which several variables represent the reported
> frequency with which participants do or use something. For example:
>
> How regularly do you use the telephone for communication with your peers?
> Never (1)
> Rarely (2)
> Sometimes (3)
> Often (4)
> Very Frequently (5)
>
> I want to measure if there is a significant difference in the frequencies
> reported between 2 (or more) such variables (e.g. is there a sig. dif
> in the
> frequency the telephone is used vs. email) and ideally the direction
> of the
> difference. The variables being compared use the same never-very frequently
> 1-5 ranking.
>
> What tests would people on this list recommend using to do this?
> Thanks for
> any help you can provide.
> -Rich
>
> =====================
> 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

Luca Meyer-3

R: Testing for Sig. Dif. between 2 Ordinal Variables

Chi-square test is what I would use too and then manually check for
directions when the test turns out to be significant. Just be careful about
minimum expected cell count (n>5 is advisable) to get correct chi-square
value.
Cheers,
Luca

Luca MEYER
www.lucameyer.com

-----Messaggio originale-----
Da: SPSSX(r) Discussion [mailto:[hidden email]] Per conto di David
Greenberg
Inviato: martedì 3 marzo 2009 23.40
A: [hidden email]
Oggetto: Re: Testing for Sig. Dif. between 2 Ordinal Variables

A chi-square test will do this, but it won't give you information about
direction. David Greenberg, Sociology Department, New York University

----- Original Message -----
From: Rich Goldman <[hidden email]>
Date: Tuesday, March 3, 2009 12:13 pm
Subject: Testing for Sig. Dif. between 2 Ordinal Variables
To: [hidden email]

> Hi. I have survey data in which several variables represent the
> reported frequency with which participants do or use something. For
example:

>
> How regularly do you use the telephone for communication with your peers?
> Never (1)
> Rarely (2)
> Sometimes (3)
> Often (4)
> Very Frequently (5)
>
> I want to measure if there is a significant difference in the
> frequencies reported between 2 (or more) such variables (e.g. is there
> a sig. dif in the frequency the telephone is used vs. email) and
> ideally the direction of the difference. The variables being compared
> use the same never-very frequently
> 1-5 ranking.
>
> What tests would people on this list recommend using to do this?
> Thanks for
> any help you can provide.
> -Rich
>
> =====================
> 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

=====================
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

Marta Garcia-Granero

Re: R: Testing for Sig. Dif. between 2 Ordinal Variables

I would NOT use chi-square, since it ignores the ordinal nature of the
data (and its potential pairing). You don't make it clear, but I assume
that you measure both variables (e.g. Phone vs Email) on the same
subjects (you have pairs of data). If you have paired data you should
use Wilcoxon test (or sign test if the differences between both
measures look highly skewed). If you have independent data (both
variables measures on different subjects), with a 5 point scale like the
one you have, you can use Mann-Whitney's U test. If you have more than 2
groups, replace Wilcoxon by Friedman test (and Mann-Whitney's U, in case
of independent samples, by Kruskal-Wallis test).

If you give me more details about the design of your study I can be more
explicit.

HTH,
Marta García-Granero

> ----- Original Message -----
> From: Rich Goldman <[hidden email]>
> Date: Tuesday, March 3, 2009 12:13 pm
> Subject: Testing for Sig. Dif. between 2 Ordinal Variables
> To: [hidden email]
>
>
>
>> Hi. I have survey data in which several variables represent the
>> reported frequency with which participants do or use something. For
>>
> example:
>
>> How regularly do you use the telephone for communication with your peers?
>> Never (1)
>> Rarely (2)
>> Sometimes (3)
>> Often (4)
>> Very Frequently (5)
>>
>> I want to measure if there is a significant difference in the
>> frequencies reported between 2 (or more) such variables (e.g. is there
>> a sig. dif in the frequency the telephone is used vs. email) and
>> ideally the direction of the difference. The variables being compared
>> use the same never-very frequently
>> 1-5 ranking.
>>
>> What tests would people on this list recommend using to do this?
>> Thanks for
>> any help you can provide.
>> -Rich
>>
>> =====================
>> 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
>
> =====================
> 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
>
>

Daciuk, Tim

Re: R: Testing for Sig. Dif. between 2 Ordinal Variables

I agree with Marta. It would appear that the data is paired. I was thinking that Kendall's tau may be appropriate.

For questions of this sort, take a look at http://www.microsiris.com/Statistical%20Decision%20Tree/ . This is based on the old A Guide for Selecting Statistical Techniques for Analyzing Social Science Data monograph from ISR. It is very helpful for coming up with suggested tests.

Tim Daciuk
Director, Worldwide Demo Resources
SPSS Inc.

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Marta García-Granero
Sent: Wednesday, March 04, 2009 3:04 AM
To: [hidden email]
Subject: Re: R: Testing for Sig. Dif. between 2 Ordinal Variables

I would NOT use chi-square, since it ignores the ordinal nature of the
data (and its potential pairing). You don't make it clear, but I assume
that you measure both variables (e.g. Phone vs Email) on the same
subjects (you have pairs of data). If you have paired data you should
use Wilcoxon test (or sign test if the differences between both
measures look highly skewed). If you have independent data (both
variables measures on different subjects), with a 5 point scale like the
one you have, you can use Mann-Whitney's U test. If you have more than 2
groups, replace Wilcoxon by Friedman test (and Mann-Whitney's U, in case
of independent samples, by Kruskal-Wallis test).

If you give me more details about the design of your study I can be more
explicit.

HTH,
Marta García-Granero

--
For miscellaneous statistical stuff, visit:
http://gjyp.nl/marta/

=====================
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

Marta Garcia-Granero

Re: R: Testing for Sig. Dif. between 2 Ordinal Variables

Daciuk, Tim wrote:
> I agree with Marta. It would appear that the data is paired. I was thinking that Kendall's tau may be appropriate.
>

Kendall Tau focuses on association, not differences. Rich Goldman wants
to test for differences between Phone/EMail, not the association between
the responses.

Regards,
Marta

(just a tiny rebuke, with a smile)
> For questions of this sort, take a look at http://www.microsiris.com/Statistical%20Decision%20Tree/ . This is based on the old A Guide for Selecting Statistical Techniques for Analyzing Social Science Data monograph from ISR. It is very helpful for coming up with suggested tests.
>
>
>

>
> -----Original Message-----
> From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Marta García-Granero
> Sent: Wednesday, March 04, 2009 3:04 AM
> To: [hidden email]
> Subject: Re: R: Testing for Sig. Dif. between 2 Ordinal Variables
>
> I would NOT use chi-square, since it ignores the ordinal nature of the
> data (and its potential pairing). You don't make it clear, but I assume
> that you measure both variables (e.g. Phone vs Email) on the same
> subjects (you have pairs of data). If you have paired data you should
> use Wilcoxon test (or sign test if the differences between both
> measures look highly skewed). If you have independent data (both
> variables measures on different subjects), with a 5 point scale like the
> one you have, you can use Mann-Whitney's U test. If you have more than 2
> groups, replace Wilcoxon by Friedman test (and Mann-Whitney's U, in case
> of independent samples, by Kruskal-Wallis test).
>
> If you give me more details about the design of your study I can be more
> explicit.
>
> HTH,
> Marta García-Granero
>
>
>> ----- Original Message -----
>> From: Rich Goldman <[hidden email]>
>> Date: Tuesday, March 3, 2009 12:13 pm
>> Subject: Testing for Sig. Dif. between 2 Ordinal Variables
>> To: [hidden email]
>>
>>
>>
>>
>>> Hi. I have survey data in which several variables represent the
>>> reported frequency with which participants do or use something. For
>>>
>>>
>> example:
>>
>>
>>> How regularly do you use the telephone for communication with your peers?
>>> Never (1)
>>> Rarely (2)
>>> Sometimes (3)
>>> Often (4)
>>> Very Frequently (5)
>>>
>>> I want to measure if there is a significant difference in the
>>> frequencies reported between 2 (or more) such variables (e.g. is there
>>> a sig. dif in the frequency the telephone is used vs. email) and
>>> ideally the direction of the difference. The variables being compared
>>> use the same never-very frequently
>>> 1-5 ranking.
>>>
>>> What tests would people on this list recommend using to do this?
>>> Thanks for
>>> any help you can provide.
>>> -Rich
>>>
>>> =====================
>>> 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
>>
>> =====================
>> 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
>>
>>
>>
>
>
> --
> For miscellaneous statistical stuff, visit:
> http://gjyp.nl/marta/
>
> =====================
> 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
>
>

Daciuk, Tim

Re: R: Testing for Sig. Dif. between 2 Ordinal Variables

Thanks Marta:
I missed that.

Tim Daciuk
Director, Worldwide Demo Resources
SPSS Inc.

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Marta García-Granero
Sent: Wednesday, March 04, 2009 12:33 PM
To: [hidden email]
Subject: Re: R: Testing for Sig. Dif. between 2 Ordinal Variables

Daciuk, Tim wrote:
> I agree with Marta. It would appear that the data is paired. I was thinking that Kendall's tau may be appropriate.
>

Kendall Tau focuses on association, not differences. Rich Goldman wants
to test for differences between Phone/EMail, not the association between
the responses.

Regards,
Marta

(just a tiny rebuke, with a smile)
> For questions of this sort, take a look at http://www.microsiris.com/Statistical%20Decision%20Tree/ . This is based on the old A Guide for Selecting Statistical Techniques for Analyzing Social Science Data monograph from ISR. It is very helpful for coming up with suggested tests.
>
>
>