if there is a command to do what your advisor wants, then why not just do
it? then if you've done things both ways, and it doesn't make a
difference, all the theorizing in the world won't matter; you'll still be
protected.
Gary
On Mon, 3 Dec 2007, Jarvier Marquez wrote:
Dear Gary,
Stata provides facilities to analyze complex survey data, but clarify
don't incorporate corrections for survey design (except the command
options "pweight" and "cluster").
I saw a discussion on the topic in the list. There, you argue that "If
you are estimating causal effects, then you almost surely do not need
the sample weights. The only issue is if the weights are a potential
omitted variable that is causally prior to your key causal variable and
correlated with it and the dependent variable." However, my advisor
recommends me to use the svy utilities arguing that they prevent from
bias and (in)efficiency. Is there a piece to justify why I do not need
sample weights? The same apply to clustering and stratification?
Kind regards,
Javier
--
Clarify mailing list served by Harvard-MIT Data Center
[Un]Subscribe/View Archive:
http://lists.gking.harvard.edu/?info=clarify