Peter Boucher <[EMAIL PROTECTED]> writes:

> Using data from
> http://149.101.22.40/dataonline/Search/Crime/State/statebystaterun.cfm?s
> tateid=9
> (you may need to cut-and-paste both lines of the above URL to get it to
> work), I calculated the following, the entries of which show the average
> homicide rate over the surrounding five years (e.g., the entry for 1971
> shows the average homicide rate for the period 1969-1973).


> I'm no statistician, so this method of evening out the bumps may be
> wrong-headed,

It is, I'm afraid, inappropriate.  Consider this example:

year rate average
-3   10   10
-2   10    9
-1   10    8
 0    5    7
 1    5    6
 2    5    5

The rate drops from 10 to 5 in year 0, but your moving average makes
it look like the decrease started two years earlier.

> but assuming they're not hopelessly wrong, these data show
> that Tim's charge of cherry-picking is baseless.  Lott's assertion that
> the rate started to trend down well before the law appears to be
> correct.

Lott did not use moving averages in his analysis, so your numbers are
not relevant to what he did.

--
Tim

Reply via email to