Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: Modeling Paired Longitudinal - Covariance question


From   Steven Samuels <[email protected]>
To   [email protected]
Subject   Re: st: Modeling Paired Longitudinal - Covariance question
Date   Tue, 21 Dec 2010 11:50:58 -0500

On Dec 21, 2010, at 9:05 AM, Jennifer A Hoff wrote:

"But I also have a correlation between the patient in Med_Group A that was propensity matched/paired to a patient in Med_Group B."

I don't think that's so. Your model implies that there is no group-A group-B correlation on observations made at a single time point, after conditioning on the covariates. (In fact the subject terms are also independent _within_ groups). The independence will hold true in the paired model as well, assuming that you have included all the possible covariates in the construction of the propensity score.

I note that you are implementing Ho, D, K Imai, G King, and E Stuart (2007) Matching as nonparametric preprocessing for reducing model dependence in parametric causal inference. Political Analysis, Vol. 15: 199-236.

One suggestion: Add an -ar- option for the correlation in --xtgee-, instead of accepting the less plausible default -exchangeable- option.

Steven J. Samuels
[email protected]
18 Cantine's Island
Saugerties NY 12477
USA
Voice: 845-246-0774
Fax:    206-202-4783


Original Post.

The dataset I am working with is panel data â longitudinal with
measurements in 2007,2008 and 2009.
I ran the model

xtset idnum year
xtgee outcome  $demog  $comobid_$covariates  Med_group $yearlist
$MedgroupYear, family(gamma)link(power -0.5) robust i(idnum )

Next ran the data through a propensity score matching (one to one match) on Med_Group. A variable called pair identifies which observations were
matched.  Each member of the duo is assigned the same value for the
variable pair.
I want to re-run the above model on the propensity matched subset of the
original data .  I realize the covariance structure needs adjusting for
the matched propensity pair.  I have correlations between the repeated
measures of an individual - (accounted for with the xtset statement). But
I also have a correlation between the patient in Med_Group A that was
propensity matched/paired to a patient in Med_Group B.   How do I adjust
the model to accommodate that?


---------------------------------------------------------------------------------------------------------
Jennifer Hoff Lindquist
Statistician
Health Services Research and Development
Mailing Address:
VA Medical Center
508 Fulton Street (152)
Durham, NC 27705
(919) 286-0411 Ext. 4054 (New as of 5/18/09)
[email protected]
[email protected]
------
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/


© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index