Re: (hyper)sensitivity of goodness-of-fit tests

RF... Thanks for your post.

This is the point I've tried to make, but it doesn't seem to be getting
through the fog.

Regardless of what the OP intends to do with this and all the
discussions, there is plainly a lack of fit. It's not that much work to
fit these data to an alternative model that will behave properly in
both tails and in the middle as well. <sigh>

Have a great Thanksgiving!!

OMU



Reef Fish wrote:
David Jones wrote:
Richard Ulrich wrote:
On 20 Nov 2006 07:15:59 -0800, "Old Mac User"
<chendrixstats@xxxxxxxxx> wrote:

Why would you go to so much trouble to numb the test so that it
fails
to show significance?

- The obvious next step seems to me to be to *quantify* the
amount of deviation of the fit. The Ns are huge, so the tests
are big, but does it *matter* to the OP? What is the purpose of
the fit?


The data you posted indicate a strong departure from an exponential
distribution. Just a simple plot of the data reveals this, even
without applying a Chi-sq test to verify it. IMHO this departure
is
bad enough to disallow using an exponential approximation. Why not
consider another distribution (gamma, perhaps) and move on with it?
It would be much faster to do that than to waste most of the data
just to numb the Chi-sq test. OMU

I agree with OMU, "fit something else" -- *if* the sample size was
wisely chosen to detect a bad 'fit' that matters. But if the
sample
is big because data just happens to be there... what matters next
is the OP's purpose, and how much the fit (and non-fit) matters.
Does it matter that the observed 'tail' is much fatter than
predicted?

There are trends in the deviations of the fit. The best clue about
the nature of a distribution is often in the question, "How was it
generated?" Does a reason suggest itself? For the purpose of the
OP, unmentioned so far, it *might* be enough to describe the fit,
and describe the deviations.


[rest of post included without additional comments.]




morfysster@xxxxxxxxx wrote:
Thank you very much for your responses.

What if I took random subsets of the observed data, and conducted
the goodness-of-fit tests using these smaller subsets and then
used
the average of the p-values corresponding to these tests? Would
such an approach be valid?


On Nov 17, 4:16 am, "Reef Fish" <large_nassua_grou...@xxxxxxxxx>
wrote:
morfyss...@xxxxxxxxx wrote:
I have a large amount of empirical data consisting of
interarrival
times that I believe are exponentially distributed. Looking at
the
quantile-quantile plot between the empirical and
theoretical/fitted
distribution, I see an almost perfect linear relationship.


While agreeing what what others have posted in response, I would like
to point out:

(i) the Q-Q plot should not be looked at for "an almost perfect linear
relationship", but for departures from a 1:1 line;

Good point to emphasize. In infinitessimal departure from a perfect
linear fit, but with residuals
--------------++++++++++--------------------
pattern is a significant departure. That is in fact the kind of
departure the eye can easily detect where as the analytic tests will
NOT.

Any systematic SMALL departures are equally BAD, e.g.
+++++++++--------------------++++++++++++++

or
++++++++++---------------------+++++++++++++++++------------------------

of the kind of TOO FEW or TOO MANY runs.

-- Reef Fish Bob.


(ii) there are other graphical procedures more or less specifically
designed to look for departures from an exponential distribution ..
see "log-survivor plots" and/or "log-survival plots" for example;

(iii) there is an extensive literature on survival analysis /
reliability / inter-arrival times that contain various well-understood
alternatives to the exponential. As Richard has pointed out, context
is important, and it may be that other similar applications have
already homed in on a suitable alternative for the case here.

(iv) other explanations of apparent departures from the null
hypothesis in large samples arise from some other aspect of the null
hypothesis not holding: for example the data may not consist of
statistically independent values, or the data may not arise from a
fixed distribution, for example if there are seasonal/time-within-day
effects that are not being modelled: again context is important here.

David Jones

.

Relevant Pages

  • Re: (hyper)sensitivity of goodness-of-fit tests
    ... amount of deviation of the fit. ... The data you posted indicate a strong departure from an exponential ... consider another distribution and move on with it? ... There are trends in the deviations of the fit. ...
    (sci.stat.math)
  • Re: Probit analysis
    ... In a log-likelihood ratio test, you would fit the probit model ... compared with a chi-square distribution. ...
    (sci.stat.math)
  • Re: Assessing credibility of a q-q plot by presence of outliers
    ... distribution in order to reject the hypothesis that this is the correct ... The idea behind q-q plot is that you EYE can detect many ... That would be a very definite departure from the reference ...
    (sci.stat.math)
  • Re: Goodness of fitting of a distribution
    ... plot that points out that the best distribution that fit my data is a ... linear combination of a weibull and a normal distribution. ... I don't need to read your Berkeley Symposium to know that the K-S ... it is the large number of bins which reduces the ...
    (sci.stat.math)
  • Re: Goodness of fitting of a distribution
    ... plot that points out that the best distribution that fit my data is a ... linear combination of a weibull and a normal distribution. ... It is the chi-squared test with many classes which has ... it is the large number of bins which reduces the ...
    (sci.stat.math)