Questions regarding SPSS

[Yu-Ping Chang]

Hi everyone,



I need suggestions for a technical question. I have a longitudinal data set
with three time measures. The purpose of this study was to evaluate the
effect of an educational training program. The sample size is 148. Three
time measures including the demographic information are all in one file
(SPSS).Participants were only recognized by a group id (one, two and three)
but there is no id for each participant. Therefore, I can only compare group
effects but cannot analyze individual change longitudinally. All
participants filled out demographic information every time. I am thinking to
match each participant by matching their demographic information, such as
marital status, religion, age, and work units. I tried to do this manually
but it was very time-consuming and inefficient. I would like to have your
suggestions. Any advice is greatly appreciated.



Thank you.



Karen Chang

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

[Barnett, Adrian (DECS)]

Hi Karen
What you have got is a probabilistic record linkage task and there is
software for the job. Most of it is not free, but there are some which
are free. The CDC has a very basic freebie but it isn't general-purpose
if I remember correctly - that is, it has its own ideas about what types
of information it will use to make the links.

One I am familiar with, Linkage Wiz - www.linkagewiz.com  - does allow
you to specify your own variables. It's not freeware, but it does have a
free version which will handle a limited number of cases (I think it is
750 cases, which would be sufficient for your task).

If you don't have name, date of birth, or gender, the key variables for
this kind of task, your links will be uncertain (although probabilistic
record linkage does not guarantee perfect links - hence 'being called
'probabilistic', nevertheless your confidence in the results when you do
have these kinds of variables can be much greater).

One of the key concepts in setting up a record linkage run is the
weights given to each of the variables used in the process. Each of
these has two weights, one for when there is a match, and another
negative one for when it does not match. They do not have to be the same
- for instance, a matching phone number would give you more confidence
that you have a good match than a mismatch would give you that you have
two different people (because they may have changed their number from
time 1 to time 2).

Linkage Wiz provides you with a set of pretty well tried-and-tested
weights for matches and mismatches on the really key identifying
variables of first name, second name, family name, sex, date of birth,
address and social security number. However you would have to make up
your own for the variables you have, and so you'd need to think about
them pretty carefully.

Anyway, LinkageWiz isn't the only thing out there so you may prefer
something else. However if you do decide to use it, I'd recommend going
through the demonstrations on the website, which show you how to carry
out each step in a record linkage run, as the demos are a bit easier to
follow than the manual.

Regards

Adrian Barnett

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Yu-Ping Chang
Sent: Friday, 28 March 2008 5:32 AM
To: [hidden email]
Subject: Questions regarding SPSS

Hi everyone,



I need suggestions for a technical question. I have a longitudinal data
set
with three time measures. The purpose of this study was to evaluate the
effect of an educational training program. The sample size is 148. Three
time measures including the demographic information are all in one file
(SPSS).Participants were only recognized by a group id (one, two and
three)
but there is no id for each participant. Therefore, I can only compare
group
effects but cannot analyze individual change longitudinally. All
participants filled out demographic information every time. I am
thinking to
match each participant by matching their demographic information, such
as
marital status, religion, age, and work units. I tried to do this
manually
but it was very time-consuming and inefficient. I would like to have
your
suggestions. Any advice is greatly appreciated.



Thank you.



Karen Chang

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

[Dennis Deck]

I agree with Adrian that matching software should help.
Besides the package he mentions there are other free programs available
such as
Link Plus (see CDC web site) and Link King (requires SAS)


Dennis Deck, PhD
RMC Research Corporation
111 SW Columbia Street, Suite 1200
Portland, Oregon 97201-5843
voice: 503-223-8248 x715
voice: 800-788-1887 x715
fax:  [phone]
[hidden email]


-----Original Message-----
From: Yu-Ping Chang [mailto:[hidden email]]
Sent: Thursday, March 27, 2008 12:02 PM
Subject: Questions regarding SPSS

Hi everyone,



I need suggestions for a technical question. I have a longitudinal data
set with three time measures. The purpose of this study was to evaluate
the effect of an educational training program. The sample size is 148.
Three time measures including the demographic information are all in one
file (SPSS).Participants were only recognized by a group id (one, two
and three) but there is no id for each participant. Therefore, I can
only compare group effects but cannot analyze individual change
longitudinally. All participants filled out demographic information
every time. I am thinking to match each participant by matching their
demographic information, such as marital status, religion, age, and work
units. I tried to do this manually but it was very time-consuming and
inefficient. I would like to have your suggestions. Any advice is
greatly appreciated.



Thank you.



Karen Chang

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