Creating a Control Group Ad-Hoc

[Adam Martin-7]

Hi all

I have encountered a frustrating situation. Normal mantra is that you create your control and test groups beforehand (making sure they are homogeneous across the chosen variables - this is straightforward).

But can you advise me on creating control groups to match test groups that have already been created? I'm looking for a smart way (ala syntax) to do it rather than the tedious way of deleting cases by case until my groups match.

Any advice?

Thanks,
Adam

--
Cell: [phone]
Website: http://www.sigmasurveys.co.za
Blog: http://www.sigmasurveys.co.za/resources

[Maguin, Eugene]

Adam,

I'm assuming that you already have picked your matching variables. I'm also
assuming that you want to match each record in the test group (the 'cases')
with one or more records in the control group. (I assume that you do not
want to simply have a control group that is not different from the test
group in aggregate terms.)

There are ways to do the matching by syntax. Those ways are purely a custom
operation in that you write the syntax to accommodate your specific
circumstances. (To get somebody to write that code for you will require that
you provide much, much more specific details.) However, I believe that spss
has published on their website a python-based routine that performs a
case-control matching operation. Look on the spss website and also look in
the listserv archives as I know the python routine has been discussed there
several times in recent months. Try searching for 'casecontrol' or
'matching' or 'case' or 'control'. You might also look for listserv posts
from Jon Peck because I think he is the spss person who has posted about
this routine.

Gene Maguin


>>I have encountered a frustrating situation. Normal mantra is that you
create your control and test groups beforehand (making sure they are
homogeneous across the chosen variables - this is straightforward).

But can you advise me on creating control groups to match test groups that
have already been created? I'm looking for a smart way (ala syntax) to do it
rather than the tedious way of deleting cases by case until my groups match.

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

[Peck, Jon]

Yes, look for the CASECTRL extension command module on Developer Central, www.spss.com/devcentral.  The Case Control with Fuzzy Matching is the newer and better version.  It will match the controls for you based on specified linking variables with an optional tolerance.

There are several options for the dataset output.  The choice will depend on how you want to use the results.  You can just add the id(s) of the selected control to the experiment (demander) dataset.  You can construct a new dataset containing the controls along with the id of the demander case.  You can copy selected variables from the control dataset matched case to the demander dataset.

This command requires at least SPSS version 16 and the Python plugin, but no knowledge of Python is required to use it.

HTH,
Jon Peck

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Gene Maguin
Sent: Monday, February 23, 2009 8:06 AM
To: [hidden email]
Subject: Re: [SPSSX-L] Creating a Control Group Ad-Hoc

Adam,

I'm assuming that you already have picked your matching variables. I'm also
assuming that you want to match each record in the test group (the 'cases')
with one or more records in the control group. (I assume that you do not
want to simply have a control group that is not different from the test
group in aggregate terms.)

There are ways to do the matching by syntax. Those ways are purely a custom
operation in that you write the syntax to accommodate your specific
circumstances. (To get somebody to write that code for you will require that
you provide much, much more specific details.) However, I believe that spss
has published on their website a python-based routine that performs a
case-control matching operation. Look on the spss website and also look in
the listserv archives as I know the python routine has been discussed there
several times in recent months. Try searching for 'casecontrol' or
'matching' or 'case' or 'control'. You might also look for listserv posts
from Jon Peck because I think he is the spss person who has posted about
this routine.

Gene Maguin


>>I have encountered a frustrating situation. Normal mantra is that you
create your control and test groups beforehand (making sure they are
homogeneous across the chosen variables - this is straightforward).

But can you advise me on creating control groups to match test groups that
have already been created? I'm looking for a smart way (ala syntax) to do it
rather than the tedious way of deleting cases by case until my groups match.

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