Need assistance - Matching case-control SPSS

[arata]

My colleague and have have a data set of 245 cases serving as experimental/treatment sample. We have access to a non experimental/control data set (N=1700). We'd like to create a control sample by matching (1:1) on five factors, age, race etc.

I've been reading the discussions around this process here and on other listservs. I am admittedly a very, very basic syntax user (mostly around statistical analysis, not data management) and am very unfamiliar with what I am reading (python etc).

I've just come across extensions from Developer Central, but once again, I am in very unfamiliar territory, and am quite stuck.

I would appreciate any help at all. Is there are straightforward process for a novice syntax user?

Tara

[Karabiner]

If you have 1700 cases for the control group, why throw 1455 of them away?
Sounds like a tremendous loss of power. Moreover, it is very doubtful that you
will be able to reach perfect 1:1 matching with 5 variables. What about using
all 1700 non-experimental cases while controlling for the 5 matching variables
by using them as covariates?

Regards

K

[arata]

I've asked the same question about the one-to-one match and my colleague will consider drawing a larger control sample.

Does anyone know how to write the SPS syntax to draw a matched sample?

[peckjonk]

I expect that the original problem has gone away, but for others who may see this, SPSS Statistics has an extension command named FUZZY available from the SPSS Community website (www.ibm.com/developerworks/spssdevcentral) for case-control matching. It provides both exact matches on a set of control variables and fuzzy matches where the user specifies the tolerance for each match variable. You can request more than one match per case if desired. Here is a syntax example. The main input is the names of the demander (case) and supplier (control) datasets and the BY variables specifying what to match on. This example also includes the tolerance (fuzz) for each variable. The remaining specifications control where the output goes and some statistical information.

FUZZY DEMANDERDS=demander SUPPLIERDS=supplier
BY=origin cylinder SUPPLIERID=id
NEWDEMANDERIDVARS=matchedcaseid
COPYTODEMANDER=mpg accel weight
DS3=dsextra DEMANDERID=demanderid
fuzz = 0 3 DRAWPOOLSIZE=howmany.


This command requires the Python Extensions, available from the Community website.

[Ghalibnet]

Hi Peckjonk,
I tried the syntax you have in your thread but it didn't work! May be you will be able to help if I explain what I need:
I have two datasets (DataSet1 , DataSet2). DataSet1 has the cases and DataSet2 has the controls.
I want to draw controls based on one variable (DIAMETER), which is a continuos variable present in both datasets.
I want to get at least five controls per case.
I want to get fuzzy matches with 0.25 tolerance.
I downloaded Fuzzy and Python extensions.
I would be so grateful if you could send me a syntax that I can try.
Thanks.


This command requires the Python Extensions, available from the Community website.[/QUOTE]