Help with a data analysis question

[Francis]

Dear Listserv members,
I'd like to request your advice, and to thank you in advance for it. I have
conducted a study in which we collected data to evaluate the impact of a
project which worked with community-based organizations to deliver care for
HIV infected individuals. I designed the study to include a sample from
individuals served by the project as a treatment group and and those that
were not served by the project as a control. I have also conducted
propensity score matching to adjust for difference in covariates of the two
groups. A major question I have is what the ideal inferential analysis
method would be appropriate. I have two major types of variables:
dichotomous outcome variables, and ranked variables on a scale of 1-5
measuring improvements in specific intervention areas. I was considering
using Wilcoxon for the rank variables  comparing the means by the group
category (treatment/control) and McNemar to compare means of the treatment
and control group on key outcomes, but the fact that data were not
collected at baseline, only at teh end of the project makes me wonder if
this would sill be a valid approach. Can you please advice?

Thanks

Francis

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

[J P-6]

Francis,

 

Maybe this will help:

http://www.socialresearchmethods.net/kb/statsimp.php

If the study is still active you *might* consider adding a post-then-pre component:

http://www.citra.org/Assets/documents/evaluation%20design.pdf

 

Do you have any qualitative data?

 

How were the participants assigned to their respective groups?

 

Hope this is helpful.

 

Best,

John

---

From: Francis <[hidden email]>
To: [hidden email]
Sent: Thu, September 16, 2010 9:10:20 AM
Subject: Help with a data analysis question

Dear Listserv members,
I'd like to request your advice, and to thank you in advance for it. I have
conducted a study in which we collected data to evaluate the impact of a
project which worked with community-based organizations to deliver care for
HIV infected individuals. I designed the study to include a sample from
individuals served by the project as a treatment group and and those that
were not served by the project as a control. I have also conducted
propensity score matching to adjust for difference in covariates of the two
groups. A major question I have is what the ideal inferential analysis
method would be appropriate. I have two major types of variables:
dichotomous outcome variables, and ranked variables on a scale of 1-5
measuring improvements in specific intervention areas. I was considering
using Wilcoxon for the rank variables  comparing the means by the group
category (treatment/control) and McNemar to compare means of the treatment
and control group on key outcomes, but the fact that data were not
collected at baseline, only at teh end of the project makes me wonder if
this would sill be a valid approach. Can you please advice?

Thanks

Francis

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