Hi All-
I am stuck.
I am creating inverse probability of treatment weights (IPTW) by coverting
the propensity scores obtained from a logistic regression run.
For the Treatment group: IPTW = 1 / Propensity score
For the Control group: IPTW = 1 / (1 - Propensity score)
When I weight the data by the IPTWs my treatment group N rises from 69 cases
to 380 cases; my control group grows from an original 300 to 380 cases. Is
this possible? Do I need to get the Ns back to their unweighted values? If
so, how?
Also, I was told by my local university statistician that I can weight the
data by IPTWs and analyze the data (e.g., independent groups t-tests)
without worrying about adjusting the standard errors. Does this seem
reasonable?
Thanks in advance for any & all advice,
Anton
=====================
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
Locked
