--- In [email protected], "authfriend" <[EMAIL PROTECTED]> wrote:
>
> --- In [email protected], "markmeredith2002" 
> <[EMAIL PROTECTED]> wrote:
> >
> > > > > If the DC crime rate was
> > > > > relatively flat during the 90s, maybe that's an ok 
> methodology.  But
> > > > > metro cities throughout the nation experienced a dramatic 
> drop in
> > > > > crime rate starting around 1992-1993 and continuing for 
> several yrs
> > > > > and therefore the study can't prove its point w/o controlling 
> for 
> > > > > this major factor.
> > > > 
> > > > They did, by "predicting" what the crime rate *would* have
> > > > been for that period that year on the basis of the previous
> > > > five-year trend.  It's true that there might have been
> > > > *somewhat* less of a reduction if the crime rate had started
> > > > going down in early 1993, but you would have no reason to
> > > > see the sharp, sudden drop they measured during the project
> > > > on the basis of the decline you're talking about (much less
> > > > the return to "normal" a few weeks after the study).
> > 
> > The 5 yr trend is meaningless - the trend for violent crime was
> > significantly up during the 80s and then it unexpectedly and
> > dramatically turned down in the 90s, then flattened out near the end
> > of that decade.  All sorts of studies came out in the 90s supposedly
> > proving that this or that particular program was reducing crime in
> > this or that city, but in retrospect we now know that crime was
> > going down in all large cities, even ones not doing this or that.
> 
> Wow.  You're still not getting it, and I'm not sure
> how to clear it up for you.  Maybe you could just try
> reading what I wrote above a little more carefully.
> Everything you just said is utterly irrelevant in the
> case of this study.

I think Mark's point is, similar to mine a month ago on this topic, if
a model to estimate (weather and other factor) normalized crime rates
for the intervention period were developed, it would have to be a
multi-year model at least five years or so -- longer would be much
better. (I used to help develop such weather normalized models, and
used them and 20-30 + years is the norm) And if such long term
normaized models were used, even if only 5 years, they would need to
include the abortion effect variables (sub-cohorts for 16-22 yr olds
with mothers who gave birth as teens, were at low income and education
levels, single, etc.) and now known crime-factors, in order to be
useful and accurate.


> If the crime rate were trending down over those five
> years, or just the last two of those years, it would
> be reflected in the "prediction" of what the crime
> rate *would have been* if the project hadn't been
> conducted during those eight weeks.


I know you (probably) don't see it because you have never done this
type of analysis (I assume), but what you just said suggests two
models: a long term model to estimate (not predict that references the
future) a weather normalized crime rate and a short term impact model.
 Normally this would NOT be a separate step and NOT two models. A
single well constructed model would control for weather, sociological
and crime-factor variabes, and the intervention variable, ALL AT THE
SAME TIME. Apparently, for reasons not clear, they did not do this.

If they predicted a normaized crime rate, then thiis is one model. To
test the impact variable, ME, there has to be a second model. 

> > I'm saying the study design needs to be revisited due to what we
now know about the unique crime trends in the 90s.
> 
> Nope, not for an eight-week World Peace Assembly.

How are you going to control for weather effects then? You need a
longer term model to do that. As well as for other crime factors. 
 
> Now, *that's* entirely reasonable.  (The data itself is
> public; 

Is it. How can I acess it?

> what you want to see is how they set up the time-
> series analysis and what the results were.)
> 
> > OF course, akasha is right that even if the statistics hold, you
> > still need more studies looking at it from different angles.
> 
> Of course.

Akasha is always right. :)

 
> 
> > So what's going to be the practical result of all these half or 3/4
> > baked M-effect studies?
> 
> Basically nothing (except that some of them are a
> little better-cooked than you think).  But they're
> pretty interesting to discuss.


Yes, as an exploratory study, they might actually show enough promise
to warrant more research. The larger research community doesn't think
so, but who knows.







------------------------ Yahoo! Groups Sponsor --------------------~--> 
Get fast access to your favorite Yahoo! Groups. Make Yahoo! your home page
http://us.click.yahoo.com/dpRU5A/wUILAA/yQLSAA/JjtolB/TM
--------------------------------------------------------------------~-> 

To subscribe, send a message to:
[EMAIL PROTECTED]

Or go to: 
http://groups.yahoo.com/group/FairfieldLife/
and click 'Join This Group!' 
Yahoo! Groups Links

<*> To visit your group on the web, go to:
    http://groups.yahoo.com/group/FairfieldLife/

<*> To unsubscribe from this group, send an email to:
    [EMAIL PROTECTED]

<*> Your use of Yahoo! Groups is subject to:
    http://docs.yahoo.com/info/terms/
 


Reply via email to