Re: st: Using AIPW for missing data purposes in RCTs?

[Steve Samuels <sjsamuels@gmail.com>]

Ariel:

Adjustment for covariates in clinical trials can correct for chance
imbalances and also reduce the variability of estimated treatment
effects (Piantadosi, 2005, pp. 470-475). Adjutment can also be useful
for creating prognostic models and for subgroup analyses (interactions
with treatment), but the latter run into the multiple-comparison
problem.

Reference: Steven Piantadosi. 2005. Clinical Trials: 
A Methodologic Perspective. Hoboken, NJ: Wiley-Interscience.

IPTW estimators, originally designed for observational stuies
are  also applied to randomized studies:

http://www.ncbi.nlm.nih.gov/pubmed/22510495

http://biostatistics.oxfordjournals.org/content/early/2012/04/17/biostatistics.kxs009.abstract

http://www4.stat.ncsu.edu/~davidian/synergytalk.pdf

Van Der Laan, Mark. 2011. Targeted Learning: Causal Inference for Observational and Experimental Data. Springer.


Stephen:

A "doubly-robust" estimator similar, but not identical to, Stata's
-teffects aipw- estimator is Mark Lunt's -dr- which can be obtained by:

. net from http://personalpages.manchester.ac.uk/staff/mark.lunt
. net install dr


Steve
sjsamuels@gmail.com




On Jun 26, 2013, at 10:22 AM, Ariel Linden, DrPH wrote:

I am perplexed by your issue here. Why would you use adjustment strategies
in an RCT? That is, unless your randomization failed? If so, you've got all
sorts of other problems on your hands...

As for adjusting for missing data, I am not sure what you are specifically
referring to? Missing data on patient characteristics, outcomes, censoring,
what? Are you planning on analyzing your data longitudinally or as a point
treatment study? If you do not describe your problem in sufficient detail,
no one will be able to give you a helpful response.

Missing data can be imputed via Stata's -mi- suite of commands, if that
makes sense in your particular situation.

As for IPTW and missing data, you are only partially correct. In the
literature (particularly James Robins work) the reference to missing data
involves handling of censoring. In short, there are two weights that are
generated, (1) the weight based on the propensity score to model probability
of receiving treatment, and (2) the weight based on the probability of being
censored. The two weights are then combined.

There is a Stata Journal paper that came out a few years ago that shows how
to perform this manually. 

Fewell Z  et al. Controlling for time-dependent confounding using marginal
structural models. Stata Journal 2004;4( 4): 402-420.

I hope this helps

Ariel

Date: Tue, 25 Jun 2013 12:17:34 +0000
From: "Stephen Kay (TRUELIFE)" <Stephen.Kay@completetruelife.com>
Subject: st: Using AIPW for missing data purposes in RCTs?

Hello,

Stata 13 has augmented inverse probability weighting, AIPW, capabilities
aimed at observational data. I don't have Stata 13 so cannot test it out.
Does anyone know if it can be used/modified  for missing data problems in
RCTs (which is one area that AIPW is used for)? I have such a trial to
analysis and it may be worth upgrading if so. Or can anyone point me to
other ado files for Stata 12 that can do so?

Many thanks,

Steve 


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/