Bivariate Correlation Variable Limit

[Hannah State-Davey]

All,

I am currently using SPSS 14 and would like to compute a Spearman's rho
correlation matrix. However, I have 129 variables (these are questionnaire
items) and there is a limit of 100 variables in SPSS for both Spearman's
and Pearson's. However, SPSS will allow you to run a Pearson's correlation
with all 129 variables through factor analysis.

Does anyone know of any way of increasing the amount of variables SPSS can
handle through a normal bivariate correlation? Or alternatively is there
some other freely available easy to use tool that will allow me to do this?

Many thanks

Hannah

=====================
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]

Hi Hannah:

> All,
>
> I am currently using SPSS 14 and would like to compute a Spearman's rho
> correlation matrix. However, I have 129 variables (these are questionnaire
> items) and there is a limit of 100 variables in SPSS for both Spearman's
> and Pearson's. However, SPSS will allow you to run a Pearson's correlation
> with all 129 variables through factor analysis.
>
> Does anyone know of any way of increasing the amount of variables SPSS can
> handle through a normal bivariate correlation? Or alternatively is there
> some other freely available easy to use tool that will allow me to do this?

A short MATRIX program will do the trick.

HTH,
Marta García-Granero

* SAMPLE DATASET: Replace by your own data *.

* Generating 129 random variables (7 points questionnaire items) *.
INPUT PROGRAM.
- VECTOR V(129).
- LOOP #I = 1 TO 10.
-  LOOP #J = 1 TO 129.
-   COMPUTE V(#J) = TRUNC(UNIFORM(7))+1.
-  END LOOP.
-  END CASE.
- END LOOP.
- END FILE.
END INPUT PROGRAM.
EXE.
FORMAT ALL (F8).

* SYNTAX STARTS HERE *.

DATASET NAME Data.
DATASET DECLARE MatrixData.
SET MXLOOPS=130.
DATASET ACTIVATE Data.

MATRIX.
* Replace "V1 TO V129" by your variable names *.
GET Data /VAR=V1 TO V129 /MISSING=OMIT.
COMPUTE n=NROW(Data).
COMPUTE k=NCOL(Data).
* Rank data to get Spearman's correlation coefficients *.
COMPUTE Ranks=MAKE(n,k,0).
LOOP i=1 TO k.
. COMPUTE Ranks(:,i)=RNKORDER(Data(:,i)).
END LOOP.
RELEASE Data,k.
* MATRIX algebra to compute a correlation matrix *.
COMPUTE Means=CSUM(Ranks)/n.
COMPUTE StdDevs=SQRT(CSSQ(Ranks)/n-Means&**2).
COMPUTE ZScores=(Ranks-MAKE(n,1,1)*Means)/(MAKE(n,1,1)*StdDevs).
COMPUTE CorrMat=T(ZScores)*ZScores/n.
* Export correlation matrix *.
MSAVE CorrMat /TYPE=CORR /OUTFILE='MatrixData' /VARIABLES=Var1 TO Var129.
END MATRIX.

DATASET ACTIVATE MatrixData.

=====================
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]

Hannah State-Davey escribió:

> I am currently using SPSS 14 and would like to compute a Spearman's rho
> correlation matrix. However, I have 129 variables (these are questionnaire
> items) and there is a limit of 100 variables in SPSS for both Spearman's
> and Pearson's. However, SPSS will allow you to run a Pearson's correlation
> with all 129 variables through factor analysis.
>
> Does anyone know of any way of increasing the amount of variables SPSS can
> handle through a normal bivariate correlation? Or alternatively is there
> some other freely available easy to use tool that will allow me to do this?
>

There is always more than a way of doing the same tasks with SPSS. Since
factor works with more than 100 variables, you can just rank your
variables and run factor to get the Spearman's correlation matrix
(that's so because Spearman's correlation coefficients are in fact
Pearson's computed on ranks):

RANK VARIABLES=V1 TO V129 /RANK INTO RV1 TO RV129 /PRINT=NO /TIES=MEAN .

DATASET DECLARE FactorMatrixData.

FACTOR
  /VARIABLES=RV1 TO RV129
  /MATRIX=OUT(COR='FactorMatrixData')
  /ANALYSIS RV1 TO RV129
  /METHOD=CORRELATION .

DATASET ACTIVATE FactorMatrixData.

Simpler than my MATRIX solution (I tend to try to solve every problem
using it, ignoring potentially better solutions).

HTH,
Marta García-granero

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

[ViAnn Beadle]

I'm just speculating here but with 129 questionnaire variables, I'd bet the
count of cases with complete data will be very small.

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Marta García-Granero
Sent: Thursday, June 19, 2008 9:28 AM
To: [hidden email]
Subject: Re: Bivariate Correlation Variable Limit - again

Hannah State-Davey escribió:
> I am currently using SPSS 14 and would like to compute a Spearman's rho
> correlation matrix. However, I have 129 variables (these are questionnaire
> items) and there is a limit of 100 variables in SPSS for both Spearman's
> and Pearson's. However, SPSS will allow you to run a Pearson's correlation
> with all 129 variables through factor analysis.
>
> Does anyone know of any way of increasing the amount of variables SPSS can
> handle through a normal bivariate correlation? Or alternatively is there
> some other freely available easy to use tool that will allow me to do
this?
>
There is always more than a way of doing the same tasks with SPSS. Since
factor works with more than 100 variables, you can just rank your
variables and run factor to get the Spearman's correlation matrix
(that's so because Spearman's correlation coefficients are in fact
Pearson's computed on ranks):

RANK VARIABLES=V1 TO V129 /RANK INTO RV1 TO RV129 /PRINT=NO /TIES=MEAN .

DATASET DECLARE FactorMatrixData.

FACTOR
  /VARIABLES=RV1 TO RV129
  /MATRIX=OUT(COR='FactorMatrixData')
  /ANALYSIS RV1 TO RV129
  /METHOD=CORRELATION .

DATASET ACTIVATE FactorMatrixData.

Simpler than my MATRIX solution (I tend to try to solve every problem
using it, ignoring potentially better solutions).

HTH,
Marta García-granero

=====================
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]

Dear all:

Sometime ago (maybe April?),  Gjalt-Jorn Ygram Peters, offered me some
hosting space for my statistical files. I uploaded then an Acrobat file
with a flow-chart of statistical methods. I have added three new files
(plus an index to help downloading them):

- Two sav files with a list of all ICD-10 disease codes (see
http://www.listserv.uga.edu/cgi-bin/wa?A2=ind0707&L=spssx-l&F=&S=&P=5435
for more details)
- A syntax file with the last  version of the accompanying code for
BMJ's "Statistics at Square one" on line book.

I will go on adding more useful files (macros mainly) from time to time,
like code for non-parametric linear regression and Bland&Altman's
"Limits Of agreement" (LOA) analysis...

This is the link to the index of all files:
http://gjyp.nl/marta/Index.html

Since I'm not an HTML programmer, you'll find the page rather
minimalistic (it was written with MS word, in fact, and saved as HTML
file), no graphics at all.

Best regards,
Marta García-Granero

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

[zstatman]

Thanks Marta

All the best
WMB
Statistical Services

============
mailto: [hidden email]
http:\\home.earthlink.net\~info.statman
============


-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Marta García-Granero
Sent: Thursday, June 19, 2008 11:42 AM
To: [hidden email]
Subject: ICD10-Codes SPSS file available (and more)

Dear all:

Sometime ago (maybe April?),  Gjalt-Jorn Ygram Peters, offered me some
hosting space for my statistical files. I uploaded then an Acrobat file with
a flow-chart of statistical methods. I have added three new files (plus an
index to help downloading them):

- Two sav files with a list of all ICD-10 disease codes (see
http://www.listserv.uga.edu/cgi-bin/wa?A2=ind0707&L=spssx-l&F=&S=&P=5435
for more details)
- A syntax file with the last  version of the accompanying code for BMJ's
"Statistics at Square one" on line book.

I will go on adding more useful files (macros mainly) from time to time,
like code for non-parametric linear regression and Bland&Altman's "Limits Of
agreement" (LOA) analysis...

This is the link to the index of all files:
http://gjyp.nl/marta/Index.html

Since I'm not an HTML programmer, you'll find the page rather minimalistic
(it was written with MS word, in fact, and saved as HTML file), no graphics
at all.

Best regards,
Marta García-Granero

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

[Richard Ristow]

At 11:42 AM 6/19/2008, Marta García-Granero wrote:

>Since I'm not an HTML programmer, you'll find
>the Web page rather minimalistic...

You don't know what a relief it is to find a Web
page that concentrates on content, rather than flashy display!

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

[Howie Harshaw-2]

Hi Folks,

A quick question about scale construction...

What is the best way to treat item non-response, or "don't know"
responses in the construction of a summed scale?  For example, if I
had a ten-item attitude scale, and a respondent selected "don't know"
for two items, what would be the best way to treat the case?   My
instinct would be to exclude that case (and all other "incomplete"
cases) from the scale as their scale score wouldn't really be
comparable to a respondent that had completed all ten items.  But that
might really diminish the sample size for the scale (I know that a
likely explanation for item non-response is a poorly worded
question...); I suppose one approach might be to sum respondents'
answers and divide by the number of scale items they completed, but I
am concerned that by doing so, I would be comparing ;Have you run
across this before?  Any thoughts?

Thank you for any guidance you can provide.

- Howie

_______________________________

Howie Harshaw
Tel: [phone]
Email: [hidden email]
Web: www.Harfolk.ca

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

[Melissa Ives]

Howie,

If the scale is a true classical scale (kappa>.70), then you can calculate the scale as
        rnd(mean.3(var1 to var8)*8).

This is for when var1-var8 are integers (eg. days, counts, etc) and it means that as long as at least 3 items were answered construct the scale as the mean of the valid answers times the number of items to put it back in to the same format as if the items were summed and all were present.

If the items represent an index (i.e. kappa<.7, they don't hang together, such as sources of stress), then you really need to have all variables present to calculate the index.

Melissa

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Howie Harshaw
Sent: Wednesday, July 02, 2008 1:13 PM
To: [hidden email]
Subject: [SPSSX-L] Item non-response in summed scale: Guidance?

Hi Folks,

A quick question about scale construction...

What is the best way to treat item non-response, or "don't know"
responses in the construction of a summed scale?  For example, if I had a ten-item attitude scale, and a respondent selected "don't know"
for two items, what would be the best way to treat the case?   My
instinct would be to exclude that case (and all other "incomplete"
cases) from the scale as their scale score wouldn't really be comparable to a respondent that had completed all ten items.  But that might really diminish the sample size for the scale (I know that a likely explanation for item non-response is a poorly worded question...); I suppose one approach might be to sum respondents'
answers and divide by the number of scale items they completed, but I am concerned that by doing so, I would be comparing ;Have you run across this before?  Any thoughts?

Thank you for any guidance you can provide.

- Howie

_______________________________

Howie Harshaw
Tel: [phone]
Email: [hidden email]
Web: www.Harfolk.ca

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

PRIVILEGED AND CONFIDENTIAL INFORMATION
This transmittal and any attachments may contain PRIVILEGED AND
CONFIDENTIAL information and is intended only for the use of the
addressee. If you are not the designated recipient, or an employee
or agent authorized to deliver such transmittals to the designated
recipient, you are hereby notified that any dissemination,
copying or publication of this transmittal is strictly prohibited. If
you have received this transmittal in error, please notify us
immediately by replying to the sender and delete this copy from your
system. You may also call us at [phone] for assistance.

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