I am attempting to help a friend analyze interrupted time series data,
and am hoping someone on the list might have advice about how to
proceed.  The data, which are police reports of hate crimes over a three
year period, seem less than ideal.  The series is organized into weekly
frequency data (a total of 156 data points).  Approximately 35 of the
weeks had some sort of hate crime incident (and one week had as many as
75), but the remaining points are all 0 (i.e., there were no hate crimes
that week).

First, I am wondering about the amount of variability required to
reliably identify an ARIMA model for the series.  In trying to model the
series, I was amazed (and a bit frightened) by how easily all the
variability was captured by either one autoregressive or one moving
average term.  In playing with fake data, it is clear it is possible to
produce autocorrelations for any series that is not a constant, even if
it has almost no variation.  I�m wondering if there are any assumptions
(or rules of thumb) about how much variation is necessary to use ARIMA.

Second, the weeks with hate crimes are not distributed evenly over time,
so the mean and standard deviation of the series are not stable.
However, the series does not appear to need to be differenced, maybe
because although both the mean and the standard deviation increase
considerably over time, most data points are still 0.  How much of a
problem is this lack of stability of the mean and standard deviation?
Is there a diagnostic I should be using to test homogeneity?

It might be easiest for people to reply directly to me at
[EMAIL PROTECTED] and I will post a summary of helpful rules of
thumb/diagnostics.  Thank you in advance for any advice/feedback!

Barbara Lehman
Claremont Graduate University




===========================================================================
This list is open to everyone.  Occasionally, less thoughtful
people send inappropriate messages.  Please DO NOT COMPLAIN TO
THE POSTMASTER about these messages because the postmaster has no
way of controlling them, and excessive complaints will result in
termination of the list.

For information about this list, including information about the
problem of inappropriate messages and information about how to
unsubscribe, please see the web page at
http://jse.stat.ncsu.edu/
===========================================================================

Reply via email to