computing individual mean values on a scale

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

computing individual mean values on a scale

Randall Stetson
Hi everyone,
Shrive et al (2006) support using an individual's mean score on answered
scale questions as replacement values for missing items in that scale.
Is there a way to pull that off in spss?

Thanks for any help,
Randy Stetson

Shrive, F.M., Stuart, H., Quan, H. et al. Dealing with missing data in a
multi-question depression scale: a comparison of imputation methods. BMC
Med Res Methodol 6, 57 (2006). https://doi.org/10.1186/1471-2288-6-57

--
------------------ASK ME HOW YOU CAN GET INVOLVED WITH NASW-NYS
------------

Randall Stetson, Ph.D., LCSW-R

NASW-NYS Region 2 Representative

Bienestar Bilingual Counseling Center, LLC
2700 Court Street
Suite 5
Syracuse, NY 13208

(315) 395-2893 (Voice)
(888) 437-6520 (fax)

_________________________________________________________________________

Confidentiality Notice

This e-mail message, including any attachments, is for the sole use of
the
intended recipient(s) and may contain confidential and privileged
information.
Any unauthorized review, use, disclosure or distribution is prohibited.  
If
you are not the intended recipient, please contact the sender by reply
e-mail
and destroy all copies of the original message.

=====================
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
Reply | Threaded
Open this post in threaded view
|

Re: computing individual mean values on a scale

Mike
It has been a while since I have done this but I believe that you want to do the
following (copied from the ver 25 Syntax Command Reference for "Vector"):

*******************************.
Example
* Replace a case’s missing values with the mean of all
nonmissing values for that case.
DATA LIST FREE /V1 V2 V3 V4 V5 V6 V7 V8.

*I assume you have variables like SDS1 to SDS20 which you would use instead of V1 to V8.

MISSING VALUES V1 TO V8 (99).

COMPUTE MEANSUB=MEAN(V1 TO V8).

*You may want to check the maximum number of missing values and adjust Mean.n( ) if necessary.

VECTOR V=V1 TO V8.
LOOP #I=1 TO 8.
+ DO IF MISSING (V(#I)).
+ COMPUTE V(#I)=MEANSUB.
+ END IF.
END LOOP.

BEGIN DATA
1 99 2 3 5 6 7 8
2 3 4 5 6 7 8 9
2 3 5 5 6 7 8 99
END DATA.
LIST.

# The first COMPUTE command calculates the variable MEANSUB as the mean of all nonmissing values for
each case.
# VECTOR defines the vector V with the original variables as its elements.
# For each case, the loop is executed once for each variable. The COMPUTE command within the loop is
executed only when the variable has a missing value for that case. COMPUTE replaces the missing value
with the value of MEANSUB.

Page 2144 IBM SPSS Statistics 25 Command Syntax Reference

I'm sure someone will point problems or errors.

-Mike Palij
New York University


On Tue, Mar 9, 2021 at 7:44 PM Randall Stetson <[hidden email]> wrote:
Hi everyone,
Shrive et al (2006) support using an individual's mean score on answered
scale questions as replacement values for missing items in that scale.
Is there a way to pull that off in spss?

Thanks for any help,
Randy Stetson

Shrive, F.M., Stuart, H., Quan, H. et al. Dealing with missing data in a
multi-question depression scale: a comparison of imputation methods. BMC
Med Res Methodol 6, 57 (2006). https://urldefense.proofpoint.com/v2/url?u=https-3A__doi.org_10.1186_1471-2D2288-2D6-2D57&d=DwICAg&c=slrrB7dE8n7gBJbeO0g-IQ&r=A8kXUln5f-BYIUaapBvbXA&m=6RcUIU9qQIn5jN4LBRq20OxsuQb_1veU88B_g3iaNao&s=mRpcgSa6umg0lx52LVgb-zOT51dlNtrhvgbW4NJFNak&e=

--
------------------ASK ME HOW YOU CAN GET INVOLVED WITH NASW-NYS
------------

Randall Stetson, Ph.D., LCSW-R

NASW-NYS Region 2 Representative

Bienestar Bilingual Counseling Center, LLC
2700 Court Street
Suite 5
Syracuse, NY 13208

(315) 395-2893 (Voice)
(888) 437-6520 (fax)

_________________________________________________________________________

Confidentiality Notice

This e-mail message, including any attachments, is for the sole use of
the
intended recipient(s) and may contain confidential and privileged
information.
Any unauthorized review, use, disclosure or distribution is prohibited. 
If
you are not the intended recipient, please contact the sender by reply
e-mail
and destroy all copies of the original message.

=====================
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
Reply | Threaded
Open this post in threaded view
|

Re: computing individual mean values on a scale

David Marso-2
In reply to this post by Randall Stetson
How about

DO REPEAT var=variablelist.
      IF MISSING (var) var=MEAN(variablelist).
END REPEAT.

Where variablelist is your list of variables of interest.

=====================
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
Reply | Threaded
Open this post in threaded view
|

Re: computing individual mean values on a scale

Jon Peck
In reply to this post by Randall Stetson
You have gotten some syntax suggestions, but you might do better statistically using MVA or MULTIPLE IMPUTATION

On Tue, Mar 9, 2021 at 5:45 PM Randall Stetson <[hidden email]> wrote:
Hi everyone,
Shrive et al (2006) support using an individual's mean score on answered
scale questions as replacement values for missing items in that scale.
Is there a way to pull that off in spss?

Thanks for any help,
Randy Stetson

Shrive, F.M., Stuart, H., Quan, H. et al. Dealing with missing data in a
multi-question depression scale: a comparison of imputation methods. BMC
Med Res Methodol 6, 57 (2006). https://doi.org/10.1186/1471-2288-6-57

--
------------------ASK ME HOW YOU CAN GET INVOLVED WITH NASW-NYS
------------

Randall Stetson, Ph.D., LCSW-R

NASW-NYS Region 2 Representative

Bienestar Bilingual Counseling Center, LLC
2700 Court Street
Suite 5
Syracuse, NY 13208

(315) 395-2893 (Voice)
(888) 437-6520 (fax)

_________________________________________________________________________

Confidentiality Notice

This e-mail message, including any attachments, is for the sole use of
the
intended recipient(s) and may contain confidential and privileged
information.
Any unauthorized review, use, disclosure or distribution is prohibited. 
If
you are not the intended recipient, please contact the sender by reply
e-mail
and destroy all copies of the original message.

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


--
Jon K Peck
[hidden email]

===================== 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
Reply | Threaded
Open this post in threaded view
|

Re: computing individual mean values on a scale

Mike
In reply to this post by David Marso-2
Your example is certainly more concise but consider the following:

(1)  Given that there are 20 SDS items, I think one would want to make a
decision about maximum number of missing item responses -- I don't know
if there are any rules of thumb for this but I would argue that maybe one
should have at least 75% (15 items) of the responses present.  One might
want to be more or less conservative but one should keep in mind that
using mean substitution reduces the variance/standard deviation for items
and the totals.

(2)  The article points out that Multiple Imputation is still better than
mean substitution and I would suggest that the OP do both to see
what the descriptive statistics look like for the two methods as well as
for the statistical analyses (especially if it is correlational/regression).

Just my 2 cents.

-Mike Palij
New York University

On Tue, Mar 9, 2021 at 9:10 PM David Marso <[hidden email]> wrote:
How about

DO REPEAT var=variablelist.
      IF MISSING (var) var=MEAN(variablelist).
END REPEAT.

Where variablelist is your list of variables of interest.

===================== 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
Reply | Threaded
Open this post in threaded view
|

Re: computing individual mean values on a scale

Art Kendall
In reply to this post by Randall Stetson
Reading between the lines I believe that you want a summative scale score
like this
compute MyScore = mean.5 (Item01 to Item07).




Why are there missing values on scale items? Are you sure there are no
substantive reasons?


A summative scale score using a sum function correlates 1.00 with a
summative scale score using a mean function. (Within the limits of numerical
accuracy of real numbers.)
A summative scale score using a mean function looks like the item response
scale.

A summative scale score using a sum  function looks like what other
researchers may have reported.

To provide for occasional randomly missing items (Poor administration?) and
to make the cognitive link to the item response scale, I usually suggest
that new scales use the mean of items.

IF there is an existing convention for a total-type score, the mean-type
score can be adjusted.






-----
Art Kendall
Social Research Consultants
--
Sent from: http://spssx-discussion.1045642.n5.nabble.com/

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

Re: computing individual mean values on a scale

Art Kendall
It appears that there were several responses being made simultaneously.

Definitely explore why items have missing values. How frequently are item
values missing? Do there appear to be patterns of missingness?  Answers to
these questions are critical to support your reasoning process.

if you do multiple imputations only do tem on one set of variables.  IV
variables ==> IV variables, and DV variables ==> DV variables.
NOT DV variables ==> IV variables or IV ==> DV variables.

Certainly, see if it makes any difference how you handle missing values on
summative scale items. Try several ways.





-----
Art Kendall
Social Research Consultants
--
Sent from: http://spssx-discussion.1045642.n5.nabble.com/

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

Re: computing individual mean values on a scale

spss.giesel@yahoo.de
In reply to this post by Randall Stetson
Hi, Randy,

it could be done like in this example. All you need is change the arguments of the macro call.

DATA LIST /a 1 b 2 c 3 d 4 e 5.
BEGIN DATA
12345
2345 
345 
45
5
 1234
  123
   12
    1
END DATA.

* B. Macro definition.

DEFINE @mv(!POS !CMDEND)
PRESERVE.
SET PRINTBACK = ON MPRINT = ON.

!DO !old !IN (!1)
    !LET !new = !CONCAT('rec_', !old)
    COMPUTE !new = !old.
    APPLY DICTIONARY FROM * /SOURCE VAR = !old /TARGET VAR = !new.
    FORMATS !new (F12.2).
    IF (MISSING(!new)) !new = MEAN.1(
    !DO !v !IN (!1)
        !IF (!v !EQ !HEAD(!1)) !THEN !CONCAT('',!v) !ELSE !CONCAT(',',!v) !IFEND
    !DOEND
    ).
!DOEND

RESTORE.
EXECUTE.
!ENDDEFINE.

* C. Macro call.

@mv a b c d e.








Am Mittwoch, 10. März 2021, 01:44:54 MEZ hat Randall Stetson <[hidden email]> Folgendes geschrieben:


Hi everyone,
Shrive et al (2006) support using an individual's mean score on answered
scale questions as replacement values for missing items in that scale.
Is there a way to pull that off in spss?

Thanks for any help,
Randy Stetson

Shrive, F.M., Stuart, H., Quan, H. et al. Dealing with missing data in a
multi-question depression scale: a comparison of imputation methods. BMC
Med Res Methodol 6, 57 (2006). https://doi.org/10.1186/1471-2288-6-57

--
------------------ASK ME HOW YOU CAN GET INVOLVED WITH NASW-NYS
------------

Randall Stetson, Ph.D., LCSW-R

NASW-NYS Region 2 Representative

Bienestar Bilingual Counseling Center, LLC
2700 Court Street
Suite 5
Syracuse, NY 13208

(315) 395-2893 (Voice)
(888) 437-6520 (fax)

_________________________________________________________________________

Confidentiality Notice

This e-mail message, including any attachments, is for the sole use of
the
intended recipient(s) and may contain confidential and privileged
information.
Any unauthorized review, use, disclosure or distribution is prohibited. 
If
you are not the intended recipient, please contact the sender by reply
e-mail
and destroy all copies of the original message.

=====================
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
Reply | Threaded
Open this post in threaded view
|

Re: computing individual mean values on a scale

bdates
Randy,

I second what Art said about looking inside the data to determine the mechanism of the missingness. If the data on a variable are missing because of the value of the variable, then that's a bigger problem than just the data being missing. So maybe check to make sure that the range of values that respondents with missing data use is similar to the range that those without missing data use. At the risk of complicating things, a good reference is https://statisticalhorizons.com/wp-content/uploads/2012/01/Milsap-Allison.pdf 
Joost van Ginkel has posted a number of solutions to missing data on this website. He has several that use a type of mean substitution. His website is below. At the bottom of the page is a list of all the syntax he's written to address missing data.


Brian





From: SPSSX(r) Discussion <[hidden email]> on behalf of [hidden email] <[hidden email]>
Sent: Wednesday, March 10, 2021 10:58 AM
To: [hidden email] <[hidden email]>
Subject: Re: computing individual mean values on a scale
 
Hi, Randy,

it could be done like in this example. All you need is change the arguments of the macro call.

DATA LIST /a 1 b 2 c 3 d 4 e 5.
BEGIN DATA
12345
2345 
345 
45
5
 1234
  123
   12
    1
END DATA.

* B. Macro definition.

DEFINE @mv(!POS !CMDEND)
PRESERVE.
SET PRINTBACK = ON MPRINT = ON.

!DO !old !IN (!1)
    !LET !new = !CONCAT('rec_', !old)
    COMPUTE !new = !old.
    APPLY DICTIONARY FROM * /SOURCE VAR = !old /TARGET VAR = !new.
    FORMATS !new (F12.2).
    IF (MISSING(!new)) !new = MEAN.1(
    !DO !v !IN (!1)
        !IF (!v !EQ !HEAD(!1)) !THEN !CONCAT('',!v) !ELSE !CONCAT(',',!v) !IFEND
    !DOEND
    ).
!DOEND

RESTORE.
EXECUTE.
!ENDDEFINE.

* C. Macro call.

@mv a b c d e.








Am Mittwoch, 10. März 2021, 01:44:54 MEZ hat Randall Stetson <[hidden email]> Folgendes geschrieben:


Hi everyone,
Shrive et al (2006) support using an individual's mean score on answered
scale questions as replacement values for missing items in that scale.
Is there a way to pull that off in spss?

Thanks for any help,
Randy Stetson

Shrive, F.M., Stuart, H., Quan, H. et al. Dealing with missing data in a
multi-question depression scale: a comparison of imputation methods. BMC
Med Res Methodol 6, 57 (2006). https://doi.org/10.1186/1471-2288-6-57

--
------------------ASK ME HOW YOU CAN GET INVOLVED WITH NASW-NYS
------------

Randall Stetson, Ph.D., LCSW-R

NASW-NYS Region 2 Representative

Bienestar Bilingual Counseling Center, LLC
2700 Court Street
Suite 5
Syracuse, NY 13208

(315) 395-2893 (Voice)
(888) 437-6520 (fax)

_________________________________________________________________________

Confidentiality Notice

This e-mail message, including any attachments, is for the sole use of
the
intended recipient(s) and may contain confidential and privileged
information.
Any unauthorized review, use, disclosure or distribution is prohibited. 
If
you are not the intended recipient, please contact the sender by reply
e-mail
and destroy all copies of the original message.

=====================
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
Reply | Threaded
Open this post in threaded view
|

Re: computing individual mean values on a scale

Bruce Weaver
Administrator
In reply to this post by Art Kendall
Following up on Art's example, if you need the prorated sum for items 1 to 7
(when there are at least 5 valid values), multiply the mean Art shows by 7
(the number of items):

COMPUTE MySum = MEAN.5(Item01 to Item07)*7.

The result is the same as if you plugged in the mean for missing items and
them computed the sum.  



Art Kendall wrote
> Reading between the lines I believe that you want a summative scale score
> like this
> compute MyScore = mean.5 (Item01 to Item07).





-----
--
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.

--
Sent from: http://spssx-discussion.1045642.n5.nabble.com/

=====================
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
--
Bruce Weaver
bweaver@lakeheadu.ca
http://sites.google.com/a/lakeheadu.ca/bweaver/

"When all else fails, RTFM."

PLEASE NOTE THE FOLLOWING: 
1. My Hotmail account is not monitored regularly. To send me an e-mail, please use the address shown above.
2. The SPSSX Discussion forum on Nabble is no longer linked to the SPSSX-L listserv administered by UGA (https://listserv.uga.edu/).
Reply | Threaded
Open this post in threaded view
|

Re: computing individual mean values on a scale

Bruce Weaver
Administrator
In reply to this post by Jon Peck
Typical examples of multiple imputation use variables that already exist.
Here, I gather we are talking about a situation where sub-scale scores
(e.g., means or sums of selected variables) are to be included as variables
in a regression model (for example).  I have never seen an example of
multiple imputation that shows how to impute values for the missing items
needed to compute those sums or means, then compute the means or sums before
finally estimating the regression model (and computing pooled parameter
estimates).  

Furthermore, suppose one only wants a sum (or mean) if there are at least 5
valid values (as in Art K's example).  For observations with fewer than 5
values, the sub-scale score will be "missing".  If that missing sub-scale
score is also imputed, it would seem there is a two-stage process of
imputation going on.  

With all of that in mind, can anyone point to an example of multiple
imputation used to first deal with missing items needed to compute sub-scale
scores, and then missing data on those (or other) variables?  I've never
seen one, and am not sure how I would approach it.  

Cheers,
Bruce



Jon Peck wrote
> You have gotten some syntax suggestions, but you might do better
> statistically using MVA or MULTIPLE IMPUTATION





-----
--
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.

--
Sent from: http://spssx-discussion.1045642.n5.nabble.com/

=====================
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
--
Bruce Weaver
bweaver@lakeheadu.ca
http://sites.google.com/a/lakeheadu.ca/bweaver/

"When all else fails, RTFM."

PLEASE NOTE THE FOLLOWING: 
1. My Hotmail account is not monitored regularly. To send me an e-mail, please use the address shown above.
2. The SPSSX Discussion forum on Nabble is no longer linked to the SPSSX-L listserv administered by UGA (https://listserv.uga.edu/).
Reply | Threaded
Open this post in threaded view
|

Re: computing individual mean values on a scale

Bruce Weaver
Administrator
This article may go some way to addressing my question below:

Gottschall, A. C., West, S. G., & Enders, C. K. (2012). A comparison of
item-level and scale-level multiple imputation for questionnaire batteries.
Multivariate Behavioral Research, 47(1), 1-25.

Now it's a matter of finding time to read it!  



Bruce Weaver wrote

> Typical examples of multiple imputation use variables that already exist.
> Here, I gather we are talking about a situation where sub-scale scores
> (e.g., means or sums of selected variables) are to be included as
> variables
> in a regression model (for example).  I have never seen an example of
> multiple imputation that shows how to impute values for the missing items
> needed to compute those sums or means, then compute the means or sums
> before
> finally estimating the regression model (and computing pooled parameter
> estimates).  
>
> Furthermore, suppose one only wants a sum (or mean) if there are at least
> 5
> valid values (as in Art K's example).  For observations with fewer than 5
> values, the sub-scale score will be "missing".  If that missing sub-scale
> score is also imputed, it would seem there is a two-stage process of
> imputation going on.  
>
> With all of that in mind, can anyone point to an example of multiple
> imputation used to first deal with missing items needed to compute
> sub-scale
> scores, and then missing data on those (or other) variables?  I've never
> seen one, and am not sure how I would approach it.  
>
> Cheers,
> Bruce
>
>
>
> Jon Peck wrote
>> You have gotten some syntax suggestions, but you might do better
>> statistically using MVA or MULTIPLE IMPUTATION
>
>
>
>
>
> -----
> --
> Bruce Weaver

> bweaver@

> 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.
>
> --
> Sent from: http://spssx-discussion.1045642.n5.nabble.com/
>
> =====================
> To manage your subscription to SPSSX-L, send a message to

> LISTSERV@.UGA

>  (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





-----
--
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.

--
Sent from: http://spssx-discussion.1045642.n5.nabble.com/

=====================
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
--
Bruce Weaver
bweaver@lakeheadu.ca
http://sites.google.com/a/lakeheadu.ca/bweaver/

"When all else fails, RTFM."

PLEASE NOTE THE FOLLOWING: 
1. My Hotmail account is not monitored regularly. To send me an e-mail, please use the address shown above.
2. The SPSSX Discussion forum on Nabble is no longer linked to the SPSSX-L listserv administered by UGA (https://listserv.uga.edu/).
Reply | Threaded
Open this post in threaded view
|

Re: computing individual mean values on a scale

Bruce Weaver
Administrator
This chapter may be helpful too:

   https://bookdown.org/mwheymans/bookmi/missing-data-in-questionnaires.html



Bruce Weaver wrote

> This article may go some way to addressing my question below:
>
> Gottschall, A. C., West, S. G., & Enders, C. K. (2012). A comparison of
> item-level and scale-level multiple imputation for questionnaire
> batteries.
> Multivariate Behavioral Research, 47(1), 1-25.
>
> Now it's a matter of finding time to read it!  
>
>
>
> Bruce Weaver wrote
>> Typical examples of multiple imputation use variables that already exist.
>> Here, I gather we are talking about a situation where sub-scale scores
>> (e.g., means or sums of selected variables) are to be included as
>> variables
>> in a regression model (for example).  I have never seen an example of
>> multiple imputation that shows how to impute values for the missing items
>> needed to compute those sums or means, then compute the means or sums
>> before
>> finally estimating the regression model (and computing pooled parameter
>> estimates).  
>>
>> Furthermore, suppose one only wants a sum (or mean) if there are at least
>> 5
>> valid values (as in Art K's example).  For observations with fewer than 5
>> values, the sub-scale score will be "missing".  If that missing sub-scale
>> score is also imputed, it would seem there is a two-stage process of
>> imputation going on.  
>>
>> With all of that in mind, can anyone point to an example of multiple
>> imputation used to first deal with missing items needed to compute
>> sub-scale
>> scores, and then missing data on those (or other) variables?  I've never
>> seen one, and am not sure how I would approach it.  
>>
>> Cheers,
>> Bruce
>>
>>
>>
>> Jon Peck wrote
>>> You have gotten some syntax suggestions, but you might do better
>>> statistically using MVA or MULTIPLE IMPUTATION
>>
>>
>>
>>
>>
>> -----
>> --
>> Bruce Weaver
>
>> bweaver@
>
>> 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.
>>
>> --
>> Sent from: http://spssx-discussion.1045642.n5.nabble.com/
>>
>> =====================
>> To manage your subscription to SPSSX-L, send a message to
>
>> LISTSERV@.UGA
>
>>  (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
>
>
>
>
>
> -----
> --
> Bruce Weaver

> bweaver@

> 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.
>
> --
> Sent from: http://spssx-discussion.1045642.n5.nabble.com/
>
> =====================
> To manage your subscription to SPSSX-L, send a message to

> LISTSERV@.UGA

>  (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





-----
--
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.

--
Sent from: http://spssx-discussion.1045642.n5.nabble.com/

=====================
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
--
Bruce Weaver
bweaver@lakeheadu.ca
http://sites.google.com/a/lakeheadu.ca/bweaver/

"When all else fails, RTFM."

PLEASE NOTE THE FOLLOWING: 
1. My Hotmail account is not monitored regularly. To send me an e-mail, please use the address shown above.
2. The SPSSX Discussion forum on Nabble is no longer linked to the SPSSX-L listserv administered by UGA (https://listserv.uga.edu/).
Reply | Threaded
Open this post in threaded view
|

Re: computing individual mean values on a scale

bdates
Here's an article, Enders being one of the authors, that summarizes the first article Bruce sent through. Basically it says that item level imputation is far more efficient than scale level imputation. In Bruce's chapter, however, there is an interesting approach in R called passive multiple imputation, which does use scale scores from each imputation together with item scores to develop the next imputation. This doesn't really help the OP unless he knows the ins and outs of R with SPSS.


Brian

From: SPSSX(r) Discussion <[hidden email]> on behalf of Bruce Weaver <[hidden email]>
Sent: Wednesday, March 10, 2021 3:31 PM
To: [hidden email] <[hidden email]>
Subject: Re: computing individual mean values on a scale
 
This chapter may be helpful too:

   https://nam12.safelinks.protection.outlook.com/?url=https%3A%2F%2Fbookdown.org%2Fmwheymans%2Fbookmi%2Fmissing-data-in-questionnaires.html&amp;data=04%7C01%7Cbdates%40SWSOL.ORG%7C4bc73810c7ef437b266a08d8e4037427%7Cecdd61640dbd4227b0986de8e52525ca%7C0%7C0%7C637510050769580094%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C1000&amp;sdata=N6yr5gnLipyD6KCpSr0lmyWjP%2FEMDjuKJrlHxRxk5FQ%3D&amp;reserved=0



Bruce Weaver wrote
> This article may go some way to addressing my question below:
>
> Gottschall, A. C., West, S. G., & Enders, C. K. (2012). A comparison of
> item-level and scale-level multiple imputation for questionnaire
> batteries.
> Multivariate Behavioral Research, 47(1), 1-25.
>
> Now it's a matter of finding time to read it! 
>
>
>
> Bruce Weaver wrote
>> Typical examples of multiple imputation use variables that already exist.
>> Here, I gather we are talking about a situation where sub-scale scores
>> (e.g., means or sums of selected variables) are to be included as
>> variables
>> in a regression model (for example).  I have never seen an example of
>> multiple imputation that shows how to impute values for the missing items
>> needed to compute those sums or means, then compute the means or sums
>> before
>> finally estimating the regression model (and computing pooled parameter
>> estimates).  
>>
>> Furthermore, suppose one only wants a sum (or mean) if there are at least
>> 5
>> valid values (as in Art K's example).  For observations with fewer than 5
>> values, the sub-scale score will be "missing".  If that missing sub-scale
>> score is also imputed, it would seem there is a two-stage process of
>> imputation going on. 
>>
>> With all of that in mind, can anyone point to an example of multiple
>> imputation used to first deal with missing items needed to compute
>> sub-scale
>> scores, and then missing data on those (or other) variables?  I've never
>> seen one, and am not sure how I would approach it. 
>>
>> Cheers,
>> Bruce
>>
>>
>>
>> Jon Peck wrote
>>> You have gotten some syntax suggestions, but you might do better
>>> statistically using MVA or MULTIPLE IMPUTATION
>>
>>
>>
>>
>>
>> -----
>> --
>> Bruce Weaver
>
>> bweaver@
>
>> https://nam12.safelinks.protection.outlook.com/?url=http%3A%2F%2Fsites.google.com%2Fa%2Flakeheadu.ca%2Fbweaver%2F&amp;data=04%7C01%7Cbdates%40SWSOL.ORG%7C4bc73810c7ef437b266a08d8e4037427%7Cecdd61640dbd4227b0986de8e52525ca%7C0%7C0%7C637510050769580094%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C1000&amp;sdata=6RBrAZNzvfMKxxXyx79RDKmQ8z7Eu%2FLChcKPj3IdWsA%3D&amp;reserved=0
>>
>> "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.
>>
>> --
>> Sent from: https://nam12.safelinks.protection.outlook.com/?url=http%3A%2F%2Fspssx-discussion.1045642.n5.nabble.com%2F&amp;data=04%7C01%7Cbdates%40SWSOL.ORG%7C4bc73810c7ef437b266a08d8e4037427%7Cecdd61640dbd4227b0986de8e52525ca%7C0%7C0%7C637510050769580094%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C1000&amp;sdata=ii2Pn9A4Ygmr14%2Fu3n3gMQxuO5eDVKI5ZVYubJ5oyIw%3D&amp;reserved=0
>>
>> =====================
>> To manage your subscription to SPSSX-L, send a message to
>
>> LISTSERV@.UGA
>
>>  (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
>
>
>
>
>
> -----
> --
> Bruce Weaver

> bweaver@

> https://nam12.safelinks.protection.outlook.com/?url=http%3A%2F%2Fsites.google.com%2Fa%2Flakeheadu.ca%2Fbweaver%2F&amp;data=04%7C01%7Cbdates%40SWSOL.ORG%7C4bc73810c7ef437b266a08d8e4037427%7Cecdd61640dbd4227b0986de8e52525ca%7C0%7C0%7C637510050769580094%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C1000&amp;sdata=6RBrAZNzvfMKxxXyx79RDKmQ8z7Eu%2FLChcKPj3IdWsA%3D&amp;reserved=0
>
> "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.
>
> --
> Sent from: https://nam12.safelinks.protection.outlook.com/?url=http%3A%2F%2Fspssx-discussion.1045642.n5.nabble.com%2F&amp;data=04%7C01%7Cbdates%40SWSOL.ORG%7C4bc73810c7ef437b266a08d8e4037427%7Cecdd61640dbd4227b0986de8e52525ca%7C0%7C0%7C637510050769580094%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C1000&amp;sdata=ii2Pn9A4Ygmr14%2Fu3n3gMQxuO5eDVKI5ZVYubJ5oyIw%3D&amp;reserved=0
>
> =====================
> To manage your subscription to SPSSX-L, send a message to

> LISTSERV@.UGA

>  (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





-----
--
Bruce Weaver
[hidden email]
https://nam12.safelinks.protection.outlook.com/?url=http%3A%2F%2Fsites.google.com%2Fa%2Flakeheadu.ca%2Fbweaver%2F&amp;data=04%7C01%7Cbdates%40SWSOL.ORG%7C4bc73810c7ef437b266a08d8e4037427%7Cecdd61640dbd4227b0986de8e52525ca%7C0%7C0%7C637510050769580094%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C1000&amp;sdata=6RBrAZNzvfMKxxXyx79RDKmQ8z7Eu%2FLChcKPj3IdWsA%3D&amp;reserved=0

"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.

--
Sent from: https://nam12.safelinks.protection.outlook.com/?url=http%3A%2F%2Fspssx-discussion.1045642.n5.nabble.com%2F&amp;data=04%7C01%7Cbdates%40SWSOL.ORG%7C4bc73810c7ef437b266a08d8e4037427%7Cecdd61640dbd4227b0986de8e52525ca%7C0%7C0%7C637510050769580094%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C1000&amp;sdata=ii2Pn9A4Ygmr14%2Fu3n3gMQxuO5eDVKI5ZVYubJ5oyIw%3D&amp;reserved=0

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