Creating a study sample matched for key variables

Moderators: statman, Analyst Techy, andris, Fierce, GerineL, Smash

Stan8219
Posts: 1
Joined: Thu Jun 13, 2013 12:53 pm

Creating a study sample matched for key variables

Postby Stan8219 » Wed Feb 12, 2014 2:15 pm

Hi folks

I'm hoping someone can advise as to how I can create a list of matched pairs for a study?

We would ike the cases to match as follows (this would be for one individual):

Programme Participant v Non-participant
Depcat = Depcat
Sex = Sex
Age_Group = Age_Group

i.e. for every participant in the programme there is a non participant who shares deprivation status, sex and age group with the participant.

Thanks in advance.

Regards

Stan
GerineL
Moderator
Posts: 1477
Joined: Tue Jun 10, 2008 4:50 pm

Re: Creating a study sample matched for key variables

Postby GerineL » Thu Feb 13, 2014 12:18 pm

What I usually do for things like this:

- save a separate spss file for every combination of your matchin variables (e.g., young males, young females, old males, old females)
- within each of these file use select cases -> random sample -> exactly ... of the first ... cases.
with odd numbers (e.g., 17 young males) I use random number generator to determine whether experimental or non-experimental condition gets one extra (e.g., i let it select a random number between 0 and 1, if >.5 it is experimental, otherwise non-experimental).

does this help?
emil14glee
Posts: 4
Joined: Fri Jun 06, 2014 3:12 pm

Re: Creating a study sample matched for key variables

Postby emil14glee » Sat Jul 05, 2014 2:31 pm

What I usually do for things like this:

- save a separate spss file for every combination of your matchin variables (e.g., young males, young females, old males, old females)
- within each of these file use select cases -> random sample -> exactly ... of the first ... cases.
with odd numbers (e.g., 17 young males) I use random number generator to determine whether experimental or non-experimental condition gets one extra (e.g., i let it select a random number between 0 and 1, if >.5 it is experimental, otherwise non-experimental).

does this help?
I think that will surely help. :) But the best thing to do to have a better result is to plan the action first.

MARKET RESEARCH
RubenGeert
Posts: 100
Joined: Mon May 19, 2014 6:06 am

Re: Creating a study sample matched for key variables

Postby RubenGeert » Sat Jul 05, 2014 3:55 pm

Presuming that your control cases are in a separate file from the experimental cases, you can do the following:

First combine the matching variables into a single group variable as described here http://www.spss-tutorials.com/combine-c ... variables/.

Next, compute a completely random variable ("COMPUTE randvar = RV.UNI(0,1).").

Then RANK randvar BY group. This creates a new variable called rrandvar. For each group, if rrandvar GT the desired frequency (derived from the file with experimental cases) the case should be deleted. Note that you can use DO REPEAT for this (see http://www.spss-tutorials.com/do-repeat/). This step leaves you with a single file holding the exact desired sample.

Finally, make sure each case is flagged as control or experimental case and merge the two files using ADD FILES (http://www.spss-tutorials.com/add-files/).

If you'd like to try this but need extra assistance, let me know and I'll send you an example syntax file that does just this.

Kind regards,

Ruben Geert van den Berg
www.spss-tutorials.com

Who is online

Users browsing this forum: No registered users and 2 guests

cron