Re: st: Regression discontinuity with interrupted time series

[Austin Nichols <austinnichols@gmail.com>]

Joshua Mitts <joshua.mitts@yale.edu>:
If I were you, I would replicate the traditional RD analysis on long
differences of different lengths, and try to interpret patterns in the
estimated effects as indicative of how effects vary with time. You
only have one instrument, and you seem to want to estimate both a jump
at time t and a difference in trend from time t on, which amounts to
two endogenous variables, treatment T and the interaction with time
Tt.  You can instrument for T and Tt with A=(Z>0) and At, using
weights to limit your sample to a vicinity of the cutoff, in a panel
regression model, with cluster-robust SE to deal with serially
correlated errors. But assumptions are less tenable and interpretation
is trickier than in the first case.

On Wed, Mar 6, 2013 at 6:19 PM, Joshua Mitts <joshua.mitts@yale.edu> wrote:
> Hi Austin,
>
> Thank you for taking the time to reply in such depth.  Let me try to
> clarify--I don't think I explained very clearly.  I'm using a fuzzy RD
> design with a numeric cutoff that determines eligibility for the
> program.  Treatment is taken by compliers at time t.  I measure
> subjects at repeated intervals prior to and following treatment.  The
> assumptions of non-manipulability of assignment variable,
> monotonicity, and exclusion are valid at time t.
>
> I could stay in the world of traditional RD and observe differences at
> t+5 minus t-5.  But that seems unduly restrictive.
*
*   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/