SPSSX Discussion

259 Variable problem: How to calculate Spearman's correlation matrix??

Classic

List

Threaded

10 messages Options

Chipulu M.

259 Variable problem: How to calculate Spearman's correlation matrix??

Dear all,

I am newbie here and have a newbie kind of problem (I think!): I am trying to calculate a spearman correlation matrix with 259 variables. However, this number exceeds the 100 allowed: Is there any way I can increase the allowed number?

Many thanks,

Max

=====================
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: 259 Variable problem: How to calculate Spearman's correlation matrix??

Hi Max

> I am newbie here and have a newbie kind of problem (I think!): I am trying to calculate a spearman correlation matrix with 259 variables. However, this number exceeds the 100 allowed: Is there any way I can increase the allowed number?
>
>
To my knowledge, the 100 variables limit can not be bypassed, but,
since Spearman correlation coefficient is simply the Pearson correlation
computed on the ranks, you can use this trick:

1) Rank your data (the 259 variables)
2) Run FACTOR (it doesn't have that limit) and ask for the correlation
matrix

* Fake dataset with 100 cases & 259 variables (replace it by your own
real data) *.
INPUT PROGRAM.
- VECTOR x(259).
- LOOP #I = 1 TO 100.
- LOOP #J = 1 TO 259.
- COMPUTE x(#J) = NORMAL(1).
- END LOOP.
- END CASE.
- END LOOP.
- END FILE.
END INPUT PROGRAM.
execute.

* First, you must rank the 259 variables *.
RANK VARIABLES=x1 TO x259 (A) /RANK /PRINT=NO.

* Now, run FACTOR and ask for the correlation matrix *.
FACTOR
/VARIABLES Rx1 TO Rx259
/ANALYSIS Rx1 TO Rx259
/PRINT CORRELATION .

The significance is not computed, sorry. If you know how to use MATRIX
programs, then you can write something to get the p-values matrix if you
need them (ask for help if you don't know how, I'll be glad to help)

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

stace swayne

question about selecting string variables

Dear list,

I have a string variable called "semester taken" (f/2003, f2004, f2005, f2006 etc.), I would like to select years f/2004 and later, but I'm having trouble getting spss to do this. Is it because this is a string variable?

All suggestions are appreciated,

Stace

Eric Langston

Re: question about selecting string variables

Stace-

There are a couple of ways to accomplish this. The first is to recode the string variable into a numeric variable – sort descending and then auto recode OR you can write some syntax like this:

Compute semester = 0 .

If (semester_taken = “f2003”) semester = 2003 .

If (semester_taken = “f2004’) semester = 2004 .

If the Semester_Taken variable always starts with an F (I assume for Fall) and then has the year follow it, this solution would also work:

Compute semester = Number(Substr(semester_taken,2,4),F4.0).

This takes the sub-string of Semester_Taken from the second position with a length of four. Then it reads the value in as a number with the assumed format to be F4.0 (4 numbers, 0 decimals). The variable that’s created will be the default

HTH,

-Eric

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of stace swayne
Sent: Tuesday, April 07, 2009 9:02 AM
To: [hidden email]
Subject: question about selecting string variables

Dear list,

All suggestions are appreciated,

Stace

ViAnn Beadle

Re: question about selecting string variables

In reply to this post by stace swayne

What is the syntax you are having problems with.

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of stace swayne
Sent: Tuesday, April 07, 2009 8:02 AM
To: [hidden email]
Subject: question about selecting string variables

Dear list,

All suggestions are appreciated,

Stace

Richard Ristow

Re: 259 Variable problem: How to calculate Spearman's correlation

In reply to this post by Chipulu M.

At 05:26 AM 4/7/2009, Chipulu M. wrote:

>I am trying to calculate a spearman correlation matrix with 259
>variables. However, this number exceeds the 100 allowed.

Marta's given you one good solution. I'd use a highly low-tech
solution. (Code not tested):

NONPAR CORR VARIABLES = V1 TO V100.
NONPAR CORR VARIABLES = V1 TO V100 WITH V101 TO V259.
NONPAR CORR VARIABLES = V101 TO V259.

I think this covers all pairs.

It does take two more data passes. I'm not sure how that compares
with Marta's MATRIX solution.

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

Jarrod Teo-2

Re: question about selecting string variables

In reply to this post by ViAnn Beadle

Dear Stace,

I hope that you know for doing a data selection, you will need to click on data-->select cases on your software.

if you place the condition as semester_taken (I hope this is how the variable name is spelt) as the following in the condition box, an error will be generated.

semester_taken=f2004 (wrong version)

The correct version should be as follows.

semester_taken='f2004' (correct version)

For string variable, the value will need to be in single or double quotes.

I hope that helps. If not, you will need to provide more details of your problems.

Cheers
Dorraj

Date: Tue, 7 Apr 2009 09:18:03 -0600
From: [hidden email]
Subject: Re: question about selecting string variables
To: [hidden email]

What is the syntax you are having problems with.

Dear list,

All suggestions are appreciated,

Stace

Make the most of what you can do on your PC and the Web, just the way you want. Windows Live

Chipulu M.

Re: 259 Variable problem: How to calculate Spearman's correlation

In reply to this post by Richard Ristow

Thanks Richard and Marta: Marta's solution worked very well!

Max

________________________________________
From: SPSSX(r) Discussion [[hidden email]] On Behalf Of Richard Ristow [[hidden email]]
Sent: 07 April 2009 19:05
To: [hidden email]
Subject: Re: 259 Variable problem: How to calculate Spearman's correlation

At 05:26 AM 4/7/2009, Chipulu M. wrote:

>I am trying to calculate a spearman correlation matrix with 259
>variables. However, this number exceeds the 100 allowed.

Marta's given you one good solution. I'd use a highly low-tech
solution. (Code not tested):

NONPAR CORR VARIABLES = V1 TO V100.
NONPAR CORR VARIABLES = V1 TO V100 WITH V101 TO V259.
NONPAR CORR VARIABLES = V101 TO V259.

I think this covers all pairs.

It does take two more data passes. I'm not sure how that compares
with Marta's MATRIX solution.

=====================
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: 259 Variable problem: How to calculate Spearman's correlation

In reply to this post by Richard Ristow

Richard Ristow wrote:
> I am trying to calculate a spearman correlation matrix with 259
> variables. However, this number exceeds the 100 allowed.
>
> Marta's given you one good solution. I'd use a highly low-tech
> solution. (Code not tested):
>
> NONPAR CORR VARIABLES = V1 TO V100.
> NONPAR CORR VARIABLES = V1 TO V100 WITH V101 TO V259.
> NONPAR CORR VARIABLES = V101 TO V259.
Hi Richard:

It seemed a good idea, but it doesn't work (your second NONPAR CORR had
to be split in two, BTW):

NONPAR CORR VARIABLES = x1 TO x100.
NONPAR CORR VARIABLES = x101 TO x200.
NONPAR CORR VARIABLES = x201 TO x259.
NONPAR CORR VARIABLES = x1 TO x100 WITH x101 TO x200.
NONPAR CORR VARIABLES = x1 TO x100 WITH x201 TO x259.
NONPAR CORR VARIABLES = x101 TO x200 WITH x201 TO x259.

The first 3 are OK, but the fourth triggers the same message (since 200
different variables are involved).

It runs OK (tested) if we work with 50-variables chunks (to involve
50+50=100 variables every time):

NONPAR CORR VARIABLES = x1 TO x50 WITH x101 TO x150.
NONPAR CORR VARIABLES = x1 TO x50 WITH x151 TO x200.
NONPAR CORR VARIABLES = x1 TO x50 WITH x201 TO x250.
NONPAR CORR VARIABLES = x1 TO x50 WITH x251 TO x259.

NONPAR CORR VARIABLES = x51 TO x100 WITH x101 TO x150.
NONPAR CORR VARIABLES = x51 TO x100 WITH x151 TO x200.
NONPAR CORR VARIABLES = x51 TO x100 WITH x201 TO x250.
NONPAR CORR VARIABLES = x51 TO x100 WITH x251 TO x259.

and so on...

Best regards,
Marta

--
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

Richard Ristow

Re: 259 Variable problem: How to calculate Spearman's correlation

At 03:25 AM 4/8/2009, Marta García-Granero wrote:

>Richard Ristow wrote:
>>
>>I'd use a highly low-tech solution. (Code not tested):
>>
>>NONPAR CORR VARIABLES = V1 TO V100.
>>NONPAR CORR VARIABLES = V1 TO V100 WITH V101 TO V259.
>>NONPAR CORR VARIABLES = V101 TO V259.
>
>It seemed a good idea, but it doesn't work (your second NONPAR CORR had
>to be split in two, BTW):
>
>NONPAR CORR VARIABLES = x1 TO x100.
>NONPAR CORR VARIABLES = x101 TO x200.
>NONPAR CORR VARIABLES = x201 TO x259.
>NONPAR CORR VARIABLES = x1 TO x100 WITH x101 TO x200.
>NONPAR CORR VARIABLES = x1 TO x100 WITH x201 TO x259.
>NONPAR CORR VARIABLES = x101 TO x200 WITH x201 TO x259.
>
>The first 3 are OK, but the fourth triggers the
>same message (since 200 different variables are involved).
>
>It runs OK (tested) if we work with 50-variables
>chunks (to involve 50+50=100 variables every time):

Thank you, Marta. And apologies.

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