--- In [email protected], Peter <[EMAIL PROTECTED]> wrote:
>
> 
> 
> --- authfriend <[EMAIL PROTECTED]> wrote:
> 
> > --- In [email protected], TurquoiseB
> > <no_reply@> wrote:
> > >
> > > --- In [email protected], "sparaig"
> > <sparaig@> wrote:
> > > >
> > > > --- In [email protected],
> > new.morning <no_reply@> 
> > wrote:
> > > > >
> > > > >
> >
> http://istpp.org/crime_prevention/voodoo_rebuttal.html#note1
> > <snip>
> > > > > And later, he dismisses a doubling of the
> > murder rate during the
> > > > > course from 10/mo to 20/, as an "outlier".
> > Thats convenient.
> > 
> > Actually that would be 10 and 20 per week, not per
> > month.
> > 
> > > > It was an outlier within the course itself. It
> > was a one-week 
> > > > aberration due to a gang fight that saw 10
> > deaths in one 
> > > > incident, IIRC.
> > > 
> > > Yeah, new...it's perfectly legitimate to "not
> > count" an
> > > anomalous event like *that*! What are you
> > *thinking*?
> > > If you bitch about something as miniscule as
> > disregarding
> > > data because it doesn't fit the all-important
> > expectations,
> > > why you could set a precedent.
> > 
> > Actually, statistically speaking, anomalous events
> > are just that, anomalous, and may *not* be relevant
> > when one is considering longer-term trends.
> > 
> > This is from the article new morning cites:
> > 
> > Park asserts that levels of violence actually
> > increased to record 
> > levels. He confuses homicides — which accounted for
> > only 3% of 
> > violent crime in Washington during 1993 — with
> > violent crimes in 
> > general. Park asserts that the murder rate soared
> > during the 
> > experiment, and claims that "participants in the
> > project seemed 
> > serenely unaware of the mounting carnage around
> > them."
> > 
> > It is true the murder rate did not drop during the
> > course — as we 
> > acknowledged in the initial research report and in
> > the published 
> > study — but the facts were very different. For six
> > weeks ending the 
> > month before the experiment, from mid-March through
> > April, homicides 
> > in Washington averaged ten per week. Beginning one
> > week after the 
> > course and for twelve weeks thereafter, homicides
> > also averaged ten 
> > per week. During the eight weeks of the experiment,
> > in June and July, 
> > the average was again ten per week — except for one
> > horrific 36-hour 
> > period in which ten people died. Apart from this
> > brief episode, which 
> > was a statistical outlier, the level of homicides
> > during June and 
> > July of 1993 was not significantly higher than the
> > remainder of the 
> > year.
> > 
> > According to his article, Park apparently took his
> > lead on the murder 
> > issue from a Washington Post reporter who had been
> > impressed that the 
> > one 36-hour period had led to a sudden doubling of
> > the murder rate 
> > that week. The reporter, and Park, did not notice
> > that the very next 
> > week the murder rate dropped from its common rate of
> > ten by more than 
> > twice — that is, the totals went up to 20 one week
> > and down to 4 the 
> > next. This is precisely the type of sporadic
> > fluctuation one must 
> > account for when total numbers are small. The
> > average incidence of 
> > murder in Washington was little more than one per
> > day, and with 
> > numbers as low as this, as Park and all scientists
> > know, random 
> > fluctuations can appear extremely high when listed
> > as percentages.
> > 
> > As I said in an earlier post, I'd sure like
> > new morning to elucidate what he thinks is
> > wrong with this explanation of why the fact
> > that the murder rate jumped during one 36-
> > hour period should not be considered significant
> > with regard to the overall study results.
> 
> The real problem with the study is the design itself.
> If it had a better design than a simple pre-post
> (which  makes no sense for research of this sort) non
> of these question would be discussed



Where is such a thing discussed? The research took several years to complete 
because 
they wanted years of post-test-period  data, IIRC.

Reply via email to