Coding ordinal level

classic Classic list List threaded Threaded
12 messages Options
Reply | Threaded
Open this post in threaded view
|

Coding ordinal level

E. Bernardo
Hello members,

In SEM, can we use -2,-1,0,+1,+2 instead of 1,2,3,4 for a Likert type scale?  What do we call the coding format that use -2,-1,0,+1,+2?

Thank you for your help.

Eins
Reply | Threaded
Open this post in threaded view
|

Re: Coding ordinal level

Art Kendall
I have not had the opportunity to try that but do not see why any 5 equally ordered values could be used.
In terms of using that response scale in data gathering, be sure that you respondents can deal with those anchors.  I would be leery about using negative values if the respondents were children who were only familiar with the non-negative part of the number line.

Are the anchors still SD, D,  N, A, SA?
If so it could be called a "Likert response scale". (as opposed to a "Likert scale" which is a set of such items.) For most audiences, i suggest you also include the response scale in the write-up.

Are you dealing with single items to measure a construct, or are these intended to be part of a summative scale?

Why do you believe the levels are severely discrepant from interval level?



Art Kendall
Social Research Consultants
On 9/27/2013 2:53 AM, E. Bernardo [via SPSSX Discussion] wrote:
Hello members,

In SEM, can we use -2,-1,0,+1,+2 instead of 1,2,3,4 for a Likert type scale?  What do we call the coding format that use -2,-1,0,+1,+2?

Thank you for your help.

Eins



If you reply to this email, your message will be added to the discussion below:
http://spssx-discussion.1045642.n5.nabble.com/Coding-ordinal-level-tp5722263.html
To start a new topic under SPSSX Discussion, email [hidden email]
To unsubscribe from SPSSX Discussion, click here.
NAML

Art Kendall
Social Research Consultants
Reply | Threaded
Open this post in threaded view
|

Re: Coding ordinal level

Bruce Weaver
Administrator
I think Art left out a not in his first sentence below -- i.e., he does not see why any 5 equally ordered values could NOT be used.  

Just another observation:  If your original data has values 1-5, centering on 3 yields your -2 to +2 scale.  So your question boils down to is it okay to center a variable in SEM.  I'm no expert in SEM, but I see no reason why you couldn't do that.

Finally, a colleague once did a little study to see if using 1-5 vs -2 to +2 at the data collection stage made any difference.  I believe he found that it did have an effect, but I don't know the details.  

HTH.


Art Kendall wrote
I have not had the
        opportunity to try that but do not see why any 5 equally ordered
        values could be used.
        In terms of using that response scale in data gathering, be sure
        that you respondents can deal with those anchors.  I would be
        leery about using negative values if the respondents were
        children who were only familiar with the non-negative part of
        the number line.
       
        Are the anchors still SD, D,  N, A, SA? If so it could be
      called a "Likert response scale". (as opposed to a "Likert scale"
      which is a set of such items.) For most audiences, i suggest you
      also include the response scale in the write-up.
     
      Are you dealing with single items to measure a construct, or are
      these intended to be part of a summative scale?
     
      Why do you believe the levels are severely discrepant from
      interval level?
     
     
     
      Art Kendall
Social Research Consultants
      On 9/27/2013 2:53 AM, E. Bernardo [via SPSSX Discussion] wrote:
   
   
     
        Hello members,
       
       
        In SEM, can we use -2,-1,0,+1,+2 instead of 1,2,3,4 for a
          Likert type scale?  What do we call the coding format that
          use -2,-1,0,+1,+2?
       
         
        Thank you for your help.
       
         
        Eins
     
     
     
     
     
        If you reply to this email, your
          message will be added to the discussion below:
        http://spssx-discussion.1045642.n5.nabble.com/Coding-ordinal-level-tp5722263.html 
     
     
        To start a new topic under SPSSX Discussion, email
        [hidden email] 
        To unsubscribe from SPSSX Discussion, click
          here .
        NAML
--
Bruce Weaver
bweaver@lakeheadu.ca
http://sites.google.com/a/lakeheadu.ca/bweaver/

"When all else fails, RTFM."

NOTE: My Hotmail account is not monitored regularly.
To send me an e-mail, please use the address shown above.
Reply | Threaded
Open this post in threaded view
|

Re: Coding ordinal level

Art Kendall
I definitely meant NOT.

I would stick with 1 to 5 because even young children are familiar with the positive number line  (counting numbers). Later they get to include zero.  Negative parts of the number line are even difficult for some adults.

Since negative numbers are understood later in individual development perhaps their use induces the need for greater cognitive effort on the part of the respondent.

Even in cultural development zero is a relatively late idea.  One of the great historical advances was the introduction of the numeral 0 in the 5th century of the current era.

I do make an exception for extent scales.
I would use 0 to 4, or 0 to 10, i.e., zero as the low (left side) anchor.
This associates never or almost never with zero. Zero is associated with none, nothing, nothingness, non being.
Likewise for none or almost none.

I also keep low anchors on the left and high anchors on the right. [or low numbers down and high number up in a graph].

I try to keep intervals between anchors 'equal-appearing' in order to enhance their association to the equal intervals on the number line.

WRT your colleagues study:
I have a guess about why there would be a discrepancy between the two sets of anchors.  One is cognitively more complex.  I guess that the degree to which the response scale makes a difference would be negatively related to age and numeracy.
Using the easier to process response scale may or may not make a difference, but why take a chance on increasing measurement error.
Art Kendall
Social Research Consultants
On 9/27/2013 8:35 AM, Bruce Weaver [via SPSSX Discussion] wrote:
I think Art left out a not in his first sentence below -- i.e., he does not see why any 5 equally ordered values could NOT be used.  

Just another observation:  If your original data has values 1-5, centering on 3 yields your -2 to +2 scale.  So your question boils down to is it okay to center a variable in SEM.  I'm no expert in SEM, but I see no reason why you couldn't do that.

Finally, a colleague once did a little study to see if using 1-5 vs -2 to +2 at the data collection stage made any difference.  I believe he found that it did have an effect, but I don't know the details.  

HTH.


Art Kendall wrote
I have not had the
        opportunity to try that but do not see why any 5 equally ordered
        values could be used.
        In terms of using that response scale in data gathering, be sure
        that you respondents can deal with those anchors.  I would be
        leery about using negative values if the respondents were
        children who were only familiar with the non-negative part of
        the number line.
       
        Are the anchors still SD, D,  N, A, SA? If so it could be
      called a "Likert response scale". (as opposed to a "Likert scale"
      which is a set of such items.) For most audiences, i suggest you
      also include the response scale in the write-up.
     
      Are you dealing with single items to measure a construct, or are
      these intended to be part of a summative scale?
     
      Why do you believe the levels are severely discrepant from
      interval level?
     
     
     
      Art Kendall
Social Research Consultants
      On 9/27/2013 2:53 AM, E. Bernardo [via SPSSX Discussion] wrote:
   
   
     
        Hello members,
       
       
        In SEM, can we use -2,-1,0,+1,+2 instead of 1,2,3,4 for a
          Likert type scale?  What do we call the coding format that
          use -2,-1,0,+1,+2?
       
         
        Thank you for your help.
       
         
        Eins
     
     
     
     
     
        If you reply to this email, your
          message will be added to the discussion below:
        http://spssx-discussion.1045642.n5.nabble.com/Coding-ordinal-level-tp5722263.html 
     
     
        To start a new topic under SPSSX Discussion, email
        [hidden email] 
        To unsubscribe from SPSSX Discussion, click
          here .
        NAML
--
Bruce Weaver
[hidden email]
http://sites.google.com/a/lakeheadu.ca/bweaver/

"When all else fails, RTFM."

NOTE: My Hotmail account is not monitored regularly.
To send me an e-mail, please use the address shown above.



If you reply to this email, your message will be added to the discussion below:
http://spssx-discussion.1045642.n5.nabble.com/Coding-ordinal-level-tp5722263p5722276.html
To start a new topic under SPSSX Discussion, email [hidden email]
To unsubscribe from SPSSX Discussion, click here.
NAML

Art Kendall
Social Research Consultants
Reply | Threaded
Open this post in threaded view
|

Re: Coding ordinal level

David Marso
Administrator
In reply to this post by E. Bernardo
In SEM the object of analysis is typically a COVARIANCE matrix.
In what way does the subtraction of a constant alter the covariances?
-----

E. Bernardo wrote
Hello members,

In SEM, can we use -2,-1,0,+1,+2 instead of 1,2,3,4 for a Likert type scale?  What do we call the coding format that use -2,-1,0,+1,+2?

Thank you for your help.

Eins
Please reply to the list and not to my personal email.
Those desiring my consulting or training services please feel free to email me.
---
"Nolite dare sanctum canibus neque mittatis margaritas vestras ante porcos ne forte conculcent eas pedibus suis."
Cum es damnatorum possederunt porcos iens ut salire off sanguinum cliff in abyssum?"
Reply | Threaded
Open this post in threaded view
|

Re: Coding ordinal level

Rich Ulrich
In reply to this post by Art Kendall
Terminology?

I am pretty sure that the original development of Likert items did
not feature any numerical cues or clues for scoring.  I am thinking
partly of the feature than has largely disappeared -- that there
should be as many items scored in each direction (1 to 5, 5 to 1) so
that "response bias" ("the tendency to agree/disagree")  would be
eliminated automatically from the total. 
 - Before computers, people most often used their pre-defined scales for
subsequent analyses.
 - I remember taking a test or two where I am sure that responses were
distorted by unnatural and awkward wording that some "balancing" items
contained.

Does that list show "Somewhat" at both extremes?
I would say that the order of responses should be (Agree, Somewhat A., Neither
Agree or Disagree, Somewhat Disagree, Disagree)  for a "Likert item", or
"item in a Likert scale."  The dimension for Likert was Agreement.  Symmetrical
scaling for other dimensions can be call "Likert-type scaling"  for items with
5 (or however many) points.

In my experience, numbers showed up on our research scales for the
convenience of the data keypunchers.  Using 1-5 was better than 0-4 because
Fortran programs read BLANK as zero.  No one would have wanted to use the
extra columns in order to punch negatives, but the math and statistics do work
out the same, exactly, if you only subtract a constant from the scores.

But I would agree that it is likely that showing numbers to testees - however
the choices differ - is apt to have at least a tiny, subtle effect.  Showing equal
intervals, like 1-5, won't bother a researcher who has intended the scale points
to be used as equal interval. 

If you show some numbers to the testee that are other than 1-5, that should
probably be described in full; I don't know of a special name.

--
Rich Ulrich



Date: Fri, 27 Sep 2013 05:02:33 -0700
From: [hidden email]
Subject: Re: Coding ordinal level
To: [hidden email]

I have not had the opportunity to try that but do not see why any 5 equally ordered values could be used.
In terms of using that response scale in data gathering, be sure that you respondents can deal with those anchors.  I would be leery about using negative values if the respondents were children who were only familiar with the non-negative part of the number line.

Are the anchors still SD, D,  N, A, SA?
If so it could be called a "Likert response scale". (as opposed to a "Likert scale" which is a set of such items.) For most audiences, i suggest you also include the response scale in the write-up.

Are you dealing with single items to measure a construct, or are these intended to be part of a summative scale?

Why do you believe the levels are severely discrepant from interval level?



Art Kendall
Social Research Consultants
On 9/27/2013 2:53 AM, E. Bernardo [via SPSSX Discussion] wrote:
Hello members,

In SEM, can we use -2,-1,0,+1,+2 instead of 1,2,3,4 for a Likert type scale?  What do we call the coding format that use -2,-1,0,+1,+2?

Thank you for your help.

Eins

Reply | Threaded
Open this post in threaded view
|

Re: Coding ordinal level

John F Hall

Not quite Likert, but these semantic differentials (with some items reversed in polarity) didn’t use numbers: the format developed in the light of advice from Bill Belson of LSE.

 

http://surveyresearch.weebly.com/uploads/2/9/9/8/2998485/hall-brown_scale.pdf

 

There are examples using smiling/grimacing faces in the brilliantly designed interactive self-completion youth questionnaire (for ages 10 – 15) of the Understanding Society survey.

 

http://discover.ukdataservice.ac.uk/catalogue/?sn=6614&type=Data%20catalogue#documentation lists documents, including questionnaires, with hyperlinks to each.

 

The youth questionnaires are:

 

http://www.esds.ac.uk/doc/6614/mrdoc/pdf/6614_wave1_main_youth_sc_questionnaire.pdf

http://www.esds.ac.uk/doc/6614/mrdoc/pdf/6614_wave2_youth_selfcompletion_questionnaire.pdf

They are presented “back-to-front” with the happiest face on the left and the most miserable face on the right, with very small numeric codes 1 – 7 underneath and white boxes with red borders directly underneath these.  The extremes are unlabelled, but the question intro explains that:

 

“ The faces express various types of feelings.  Below each face is a number where 1 is completely happy and 7 is not at all happy.  Please tick the face closest to expressing how you feel about each of the following things.”

 

The names of the variables in the SPSS file are displayed on boxes, but I’m not sure if these were visible to the respondent.

 

Data from the series are available as SPSS or Stata files from UK Data Service.

 

 

John F Hall (Mr)

[Retired academic survey researcher]

 

Email:   [hidden email] 

Website: www.surveyresearch.weebly.com

SPSS start page:  www.surveyresearch.weebly.com/spss-without-tears.html

  

  

 

 

 

 

 

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Rich Ulrich
Sent: 27 September 2013 21:41
To: [hidden email]
Subject: Re: Coding ordinal level

 

Terminology?

I am pretty sure that the original development of Likert items did
not feature any numerical cues or clues for scoring.  I am thinking
partly of the feature than has largely disappeared -- that there
should be as many items scored in each direction (1 to 5, 5 to 1) so
that "response bias" ("the tendency to agree/disagree")  would be
eliminated automatically from the total. 
 - Before computers, people most often used their pre-defined scales for
subsequent analyses.
 - I remember taking a test or two where I am sure that responses were
distorted by unnatural and awkward wording that some "balancing" items
contained.

Does that list show "Somewhat" at both extremes?
I would say that the order of responses should be (Agree, Somewhat A., Neither
Agree or Disagree, Somewhat Disagree, Disagree)  for a "Likert item", or
"item in a Likert scale."  The dimension for Likert was Agreement.  Symmetrical
scaling for other dimensions can be call "Likert-type scaling"  for items with
5 (or however many) points.

In my experience, numbers showed up on our research scales for the
convenience of the data keypunchers.  Using 1-5 was better than 0-4 because
Fortran programs read BLANK as zero.  No one would have wanted to use the
extra columns in order to punch negatives, but the math and statistics do work
out the same, exactly, if you only subtract a constant from the scores.

But I would agree that it is likely that showing numbers to testees - however
the choices differ - is apt to have at least a tiny, subtle effect.  Showing equal
intervals, like 1-5, won't bother a researcher who has intended the scale points
to be used as equal interval. 

If you show some numbers to the testee that are other than 1-5, that should
probably be described in full; I don't know of a special name.

--
Rich Ulrich


Date: Fri, 27 Sep 2013 05:02:33 -0700
From: [hidden email]
Subject: Re: Coding ordinal level
To: [hidden email]

I have not had the opportunity to try that but do not see why any 5 equally ordered values could be used.
In terms of using that response scale in data gathering, be sure that you respondents can deal with those anchors.  I would be leery about using negative values if the respondents were children who were only familiar with the non-negative part of the number line.

Are the anchors still SD, D,  N, A, SA?
If so it could be called a "Likert response scale". (as opposed to a "Likert scale" which is a set of such items.) For most audiences, i suggest you also include the response scale in the write-up.

Are you dealing with single items to measure a construct, or are these intended to be part of a summative scale?

Why do you believe the levels are severely discrepant from interval level?


Art Kendall
Social Research Consultants

On 9/27/2013 2:53 AM, E. Bernardo [via SPSSX Discussion] wrote:

Hello members,

 

In SEM, can we use -2,-1,0,+1,+2 instead of 1,2,3,4 for a Likert type scale?  What do we call the coding format that use -2,-1,0,+1,+2?

 

Thank you for your help.

 

Eins

 

Reply | Threaded
Open this post in threaded view
|

Re: Coding ordinal level

Art Kendall
In reply to this post by Rich Ulrich
comments interspersed.
Art Kendall
Social Research Consultants
On 9/27/2013 3:42 PM, Rich Ulrich [via SPSSX Discussion] wrote:
Terminology?

I am pretty sure that the original development of Likert items did
not feature any numerical cues or clues for scoring.
IIRC Likert did not use numeric anchors.

I am thinking
partly of the feature than has largely disappeared -- that there
should be as many items scored in each direction (1 to 5, 5 to 1) so
that "response bias" ("the tendency to agree/disagree")  would be
eliminated automatically from the total. 
AFAIK having (close to) the same number of items that are used as is and those that need to have the response scale reversed is still good practice.
 - Before computers, people most often used their pre-defined scales for
subsequent analyses.
 - I remember taking a test or two where I am sure that responses were
distorted by unnatural and awkward wording that some "balancing" items
contained.
Probably true. Including double negatives, etc. Many beginners fail to do sufficient cognitive testing of their items.
However, it is even word in my opinion to alternate response scales.
some with low values on the left and some with low values on the right.

Does that list show "Somewhat" at both extremes?
I would say that the order of responses should be (Agree, Somewhat A., Neither
Agree or Disagree, Somewhat Disagree, Disagree)  for a "Likert item", or
"item in a Likert scale."  The dimension for Likert was Agreement.  Symmetrical
scaling for other dimensions can be call "Likert-type scaling"  for items with
5 (or however many) points.
Actually "approval" but has come to be "agreement".
I too like to use symmetric, it would get very weird reflecting response scales that were not symmetric.
However, because I hope to prompt tacit association with the number line, I would order the values in ascending order.

In my experience, numbers showed up on our research scales for the
convenience of the data keypunchers.  Using 1-5 was better than 0-4 because
Fortran programs read BLANK as zero.  No one would have wanted to use the
extra columns in order to punch negatives, but the math and statistics do work
out the same, exactly, if you only subtract a constant from the scores.
However, given that zero is a concept that comes late in individual development and came late in cultural development I opt for 1 to n for Likert items and 0 to n for extent items. Negative numbers also come later in individual development and came late in cultural development. Again, the effect may be subtle. But why risk the downside of increasing respondent burden?
But I would agree that it is likely that showing numbers to testees - however
the choices differ - is apt to have at least a tiny, subtle effect.  Showing equal
intervals, like 1-5, won't bother a researcher who has intended the scale points
to be used as equal interval. 
 I agree that showing both (one in the header and the other in the list of response options) might prompt association with the number line. I cannot think of a downside off hand.

If you show some numbers to the testee that are other than 1-5, that should
probably be described in full; I don't know of a special name.

--
Rich Ulrich



Date: Fri, 27 Sep 2013 05:02:33 -0700
From: [hidden email]
Subject: Re: Coding ordinal level
To: [hidden email]

I have not had the opportunity to try that but do not see why any 5 equally ordered values could be used.
In terms of using that response scale in data gathering, be sure that you respondents can deal with those anchors.  I would be leery about using negative values if the respondents were children who were only familiar with the non-negative part of the number line.

Are the anchors still SD, D,  N, A, SA?
If so it could be called a "Likert response scale". (as opposed to a "Likert scale" which is a set of such items.) For most audiences, i suggest you also include the response scale in the write-up.

Are you dealing with single items to measure a construct, or are these intended to be part of a summative scale?

Why do you believe the levels are severely discrepant from interval level?



Art Kendall
Social Research Consultants
On 9/27/2013 2:53 AM, E. Bernardo [via SPSSX Discussion] wrote:
Hello members,

In SEM, can we use -2,-1,0,+1,+2 instead of 1,2,3,4 for a Likert type scale?  What do we call the coding format that use -2,-1,0,+1,+2?

Thank you for your help.

Eins




If you reply to this email, your message will be added to the discussion below:
http://spssx-discussion.1045642.n5.nabble.com/Coding-ordinal-level-tp5722263p5722291.html
To start a new topic under SPSSX Discussion, email [hidden email]
To unsubscribe from SPSSX Discussion, click here.
NAML

Art Kendall
Social Research Consultants
Reply | Threaded
Open this post in threaded view
|

Re: Coding ordinal level

Art Kendall
In reply to this post by John F Hall
I am leery of coding the values in a response scale more to less rather than less to more.
Art Kendall
Social Research Consultants
On 9/27/2013 4:53 PM, John F Hall [via SPSSX Discussion] wrote:

Not quite Likert, but these semantic differentials (with some items reversed in polarity) didn’t use numbers: the format developed in the light of advice from Bill Belson of LSE.

 

http://surveyresearch.weebly.com/uploads/2/9/9/8/2998485/hall-brown_scale.pdf

 

There are examples using smiling/grimacing faces in the brilliantly designed interactive self-completion youth questionnaire (for ages 10 – 15) of the Understanding Society survey.

 

http://discover.ukdataservice.ac.uk/catalogue/?sn=6614&type=Data%20catalogue#documentation lists documents, including questionnaires, with hyperlinks to each.

 

The youth questionnaires are:

 

http://www.esds.ac.uk/doc/6614/mrdoc/pdf/6614_wave1_main_youth_sc_questionnaire.pdf

http://www.esds.ac.uk/doc/6614/mrdoc/pdf/6614_wave2_youth_selfcompletion_questionnaire.pdf

They are presented “back-to-front” with the happiest face on the left and the most miserable face on the right, with very small numeric codes 1 – 7 underneath and white boxes with red borders directly underneath these.  The extremes are unlabelled, but the question intro explains that:

 

“ The faces express various types of feelings.  Below each face is a number where 1 is completely happy and 7 is not at all happy.  Please tick the face closest to expressing how you feel about each of the following things.”

 

The names of the variables in the SPSS file are displayed on boxes, but I’m not sure if these were visible to the respondent.

 

Data from the series are available as SPSS or Stata files from UK Data Service.

 

 

John F Hall (Mr)

[Retired academic survey researcher]

 

Email:   [hidden email] 

Website: www.surveyresearch.weebly.com

SPSS start page:  www.surveyresearch.weebly.com/spss-without-tears.html

  

  

 

 

 

 

 

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Rich Ulrich
Sent: 27 September 2013 21:41
To: [hidden email]
Subject: Re: Coding ordinal level

 

Terminology?

I am pretty sure that the original development of Likert items did
not feature any numerical cues or clues for scoring.  I am thinking
partly of the feature than has largely disappeared -- that there
should be as many items scored in each direction (1 to 5, 5 to 1) so
that "response bias" ("the tendency to agree/disagree")  would be
eliminated automatically from the total. 
 - Before computers, people most often used their pre-defined scales for
subsequent analyses.
 - I remember taking a test or two where I am sure that responses were
distorted by unnatural and awkward wording that some "balancing" items
contained.

Does that list show "Somewhat" at both extremes?
I would say that the order of responses should be (Agree, Somewhat A., Neither
Agree or Disagree, Somewhat Disagree, Disagree)  for a "Likert item", or
"item in a Likert scale."  The dimension for Likert was Agreement.  Symmetrical
scaling for other dimensions can be call "Likert-type scaling"  for items with
5 (or however many) points.

In my experience, numbers showed up on our research scales for the
convenience of the data keypunchers.  Using 1-5 was better than 0-4 because
Fortran programs read BLANK as zero.  No one would have wanted to use the
extra columns in order to punch negatives, but the math and statistics do work
out the same, exactly, if you only subtract a constant from the scores.

But I would agree that it is likely that showing numbers to testees - however
the choices differ - is apt to have at least a tiny, subtle effect.  Showing equal
intervals, like 1-5, won't bother a researcher who has intended the scale points
to be used as equal interval. 

If you show some numbers to the testee that are other than 1-5, that should
probably be described in full; I don't know of a special name.

--
Rich Ulrich


Date: Fri, 27 Sep 2013 05:02:33 -0700
From: [hidden email]
Subject: Re: Coding ordinal level
To: [hidden email]

I have not had the opportunity to try that but do not see why any 5 equally ordered values could be used.
In terms of using that response scale in data gathering, be sure that you respondents can deal with those anchors.  I would be leery about using negative values if the respondents were children who were only familiar with the non-negative part of the number line.

Are the anchors still SD, D,  N, A, SA?
If so it could be called a "Likert response scale". (as opposed to a "Likert scale" which is a set of such items.) For most audiences, i suggest you also include the response scale in the write-up.

Are you dealing with single items to measure a construct, or are these intended to be part of a summative scale?

Why do you believe the levels are severely discrepant from interval level?


Art Kendall
Social Research Consultants

On 9/27/2013 2:53 AM, E. Bernardo [via SPSSX Discussion] wrote:

Hello members,

 

In SEM, can we use -2,-1,0,+1,+2 instead of 1,2,3,4 for a Likert type scale?  What do we call the coding format that use -2,-1,0,+1,+2?

 

Thank you for your help.

 

Eins

 




If you reply to this email, your message will be added to the discussion below:
http://spssx-discussion.1045642.n5.nabble.com/Coding-ordinal-level-tp5722263p5722292.html
To start a new topic under SPSSX Discussion, email [hidden email]
To unsubscribe from SPSSX Discussion, click here.
NAML

Art Kendall
Social Research Consultants
Reply | Threaded
Open this post in threaded view
|

Re: Coding ordinal level

Art Kendall
In reply to this post by John F Hall
I am leery of coding response scale values for extent scales  from more to less rather than less to more.
Also, very much like the problems with Chernoff faces, it takes a close look to see the differences in the curvature of the mouth in 1 vs 2 vs 3.

These surveys are already done so people will have to use the data as it is.

To be fair these studies were done before the prevalence of cognitive considerations in survey research, but I would advise against their use in any new research.
Art Kendall
Social Research Consultants
On 9/27/2013 4:53 PM, John F Hall [via SPSSX Discussion] wrote:

There are examples using smiling/grimacing faces in the brilliantly designed interactive self-completion youth questionnaire (for ages 10 – 15) of the Understanding Society survey.

 

http://discover.ukdataservice.ac.uk/catalogue/?sn=6614&type=Data%20catalogue#documentation lists documents, including questionnaires, with hyperlinks to each.

 

The youth questionnaires are:

 

http://www.esds.ac.uk/doc/6614/mrdoc/pdf/6614_wave1_main_youth_sc_questionnaire.pdf

http://www.esds.ac.uk/doc/6614/mrdoc/pdf/6614_wave2_youth_selfcompletion_questionnaire.pdf

They are presented “back-to-front” with the happiest face on the left and the most miserable face on the right, with very small numeric codes 1 – 7 underneath and white boxes with red borders directly underneath these.  The extremes are unlabelled, but the question intro explains that:

 

“ The faces express various types of feelings.  Below each face is a number where 1 is completely happy and 7 is not at all happy.  Please tick the face closest to expressing how you feel about each of the following things.”

 

The names of the variables in the SPSS file are displayed on boxes, but I’m not sure if these were visible to the respondent.

 

Data from the series are available as SPSS or Stata files from UK Data Service.

 

 

John F Hall (Mr)

[Retired academic survey researcher]

 

Email:   [hidden email] 

Website: www.surveyresearch.weebly.com

SPSS start page:  www.surveyresearch.weebly.com/spss-without-tears.html

  

  

 

 

 

 

 

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Rich Ulrich
Sent: 27 September 2013 21:41
To: [hidden email]
Subject: Re: Coding ordinal level

 

Terminology?

I am pretty sure that the original development of Likert items did
not feature any numerical cues or clues for scoring.  I am thinking
partly of the feature than has largely disappeared -- that there
should be as many items scored in each direction (1 to 5, 5 to 1) so
that "response bias" ("the tendency to agree/disagree")  would be
eliminated automatically from the total. 
 - Before computers, people most often used their pre-defined scales for
subsequent analyses.
 - I remember taking a test or two where I am sure that responses were
distorted by unnatural and awkward wording that some "balancing" items
contained.

Does that list show "Somewhat" at both extremes?
I would say that the order of responses should be (Agree, Somewhat A., Neither
Agree or Disagree, Somewhat Disagree, Disagree)  for a "Likert item", or
"item in a Likert scale."  The dimension for Likert was Agreement.  Symmetrical
scaling for other dimensions can be call "Likert-type scaling"  for items with
5 (or however many) points.

In my experience, numbers showed up on our research scales for the
convenience of the data keypunchers.  Using 1-5 was better than 0-4 because
Fortran programs read BLANK as zero.  No one would have wanted to use the
extra columns in order to punch negatives, but the math and statistics do work
out the same, exactly, if you only subtract a constant from the scores.

But I would agree that it is likely that showing numbers to testees - however
the choices differ - is apt to have at least a tiny, subtle effect.  Showing equal
intervals, like 1-5, won't bother a researcher who has intended the scale points
to be used as equal interval. 

If you show some numbers to the testee that are other than 1-5, that should
probably be described in full; I don't know of a special name.

--
Rich Ulrich


Date: Fri, 27 Sep 2013 05:02:33 -0700
From: [hidden email]
Subject: Re: Coding ordinal level
To: [hidden email]

I have not had the opportunity to try that but do not see why any 5 equally ordered values could be used.
In terms of using that response scale in data gathering, be sure that you respondents can deal with those anchors.  I would be leery about using negative values if the respondents were children who were only familiar with the non-negative part of the number line.

Are the anchors still SD, D,  N, A, SA?
If so it could be called a "Likert response scale". (as opposed to a "Likert scale" which is a set of such items.) For most audiences, i suggest you also include the response scale in the write-up.

Are you dealing with single items to measure a construct, or are these intended to be part of a summative scale?

Why do you believe the levels are severely discrepant from interval level?


Art Kendall
Social Research Consultants

On 9/27/2013 2:53 AM, E. Bernardo [via SPSSX Discussion] wrote:

Hello members,

 

In SEM, can we use -2,-1,0,+1,+2 instead of 1,2,3,4 for a Likert type scale?  What do we call the coding format that use -2,-1,0,+1,+2?

 

Thank you for your help.

 

Eins

 




If you reply to this email, your message will be added to the discussion below:
http://spssx-discussion.1045642.n5.nabble.com/Coding-ordinal-level-tp5722263p5722292.html
To start a new topic under SPSSX Discussion, email [hidden email]
To unsubscribe from SPSSX Discussion, click here.
NAML

Art Kendall
Social Research Consultants
Reply | Threaded
Open this post in threaded view
|

Re: Coding ordinal level

Robert Jones
I would just add that if the questions in a Likert scale are presented to the people participating in a survey as numeric without descriptive labels, then the results could be different if all possible responses are positive compared with when they are balanced between negative and positive.  I myself might well regard a response of 2 in a positive scale of five without descriptive labels as very mildly agreeing, whereas that would not be the case if I had to make a definitely negative response.
Reply | Threaded
Open this post in threaded view
|

Re: Coding ordinal level

Art Kendall
You are exactly right.
It would not make sense to me to have a number scale without anchors

Art Kendall
Social Research Consultants
On 9/28/2013 6:55 PM, Robert Jones [via SPSSX Discussion] wrote:
I would just add that if the questions in a Likert scale are presented to the people participating in a survey as numeric without descriptive labels, then the results could be different if all possible responses are positive compared with when they are balanced between negative and positive.  I myself might well regard a response of 2 in a positive scale of five without descriptive labels as very mildly agreeing, whereas that would not be the case if I had to make a definitely negative response.


If you reply to this email, your message will be added to the discussion below:
http://spssx-discussion.1045642.n5.nabble.com/Coding-ordinal-level-tp5722263p5722309.html
To start a new topic under SPSSX Discussion, email [hidden email]
To unsubscribe from SPSSX Discussion, click here.
NAML

Art Kendall
Social Research Consultants