cescript@mtu.edu (Charles Scripter) writes:
>
> On Fri, 9 Aug 1996 20:39:15 GMT, Enturbulated (entheta@eskimo.com) wrote:
>
> > You have yet to show that the model was bogus.
> Poor Pim, that's exactly what the sci.stat.edu folks pointed out.
> What do you think Mr. Neal meant by "This is invalid"?
This is not showing was invalid , this is claiming that the test was
invalid. Where is the statistical test for autocorrelation? Dr
McDonald did such a test but you didn't like and decided to call him
names.
> Do you think
> that this meant that he considered Tim's model correct?... Clearly
> Pim must also think that Dr. Young was supporting Tim's claim... ;)
>
> My own demonstration of Tim's error was to point out that Tim's
> model results in every year being "significant".
No it doesn't. Despite explanations in great length and detail from
me and others you continue to misrepresent what the model shows.
You have maintained that there was no reduction in homicides in NSW
following the introduction of gun control. A t-test on the difference
in the means shows a significant reduction. This result is robust,
that is, even you arbitrarily move one or two years from the before
set to the after set or vice versa the difference is still significant
(but smaller). It is truly perverse to argue that this robustness
proves that there was no reduction.
> > Your assertion that the
> > data were correlated however did not seem to be supported on the
> > sci.stat group.
> + From: radford@cs.toronto.edu (Radford Neal)
> + In article <tnohxolj3d.fsf@alboka.orchestra.cse.unsw.EDU.AU>,
> + Tim Lambert <lambert@alboka.orchestra.cse.unsw.EDU.AU> wrote:
> +
> + >Actually, my analysis method was:
> + >
> + >Mean homicide rate in years 1910-1920: 2.3
> + >Mean homicide rate in years 1921-1930: 1.5
> + >Difference = -0.8 t statistic is -4.9 (19df) p = 0.00009
> + >I.e. the mean homicide rate 1921-30 is significantly less than that
> + >for 1910-20.
> +
> + This is invalid. You are apparently viewing the homicide rates
> + in different years as being independent, but they clearly are not.
This is "asserted" Charles, not "supported". Mr Neal did not perform
any statistical tests to reject the null hypothesis of uncorrelated
data. You've seen the results of such tests and all you can do is
call Dr McDonald names.
--
Tim