Re: Rank order

On 4 Apr 2006 08:42:17 -0700, "alves" <alves.paulo@xxxxxxxxx> wrote:

Hi,

I have a dataset with 20 years and in each year 40 countries are ranked
on a variable. I need to check if the ranking are statistically the
same over the period or not. I was looking at the Kendall's Coefficient
of concordance, but in this case the H0 is that W=0, in other words
that the correlation between ranking is 0. What I need is H0=1.

Sorry if this is a silly question, but I would really appreciate some
help...

You need help in constructing the hypothesis.

Okay. You *know* that H0: W=1 is not true. Right?

It's probably not true for the rank-correlation for any
two consecutive years, not to mention 20 years, just
because of errors in measurements, every year.


Now, if there are *changes* taking place across time,
then year 1 will be more like year 2 than it is like
year 20. This suggests that you could look pair-wise at
rank-correlations at all 1-year lags, 2-year lags, and so on ...
- You might plot the averages for each amount of lag.
- You might test the correlation for (1,2) versus (1,20).

You can assess the consistency between years, or across
years, in various ways. Who are you trying to convince,
about what? How much change *is* there? What point
are you going to show, in the end?

Hope this helps.

--
Rich Ulrich, wpilib@xxxxxxxx
http://www.pitt.edu/~wpilib/index.html
.

Relevant Pages

  • Re: The Rating System Again
    ... > of the correlation between Ranking and Games Played it has a value of ... > Games Played and Ranking, which would be expected if the Tatu factor was ... more than it represents those players who played lots. ...
    (rec.games.trading-cards.jyhad)
  • Re: The Rating System Again
    ... > crunching to understand how much a ninth tournament adds to the ... The R^2 value is a measure of the correlation between two variables a ... Games Played and Ranking, which would be expected if the Tatu factor was ... more than it represents those players who played lots. ...
    (rec.games.trading-cards.jyhad)
  • Re: Using correlation coefficient
    ... >> predicted raw values, the other was the compilation of many ... > Ranking also filters out information which may be obscuring important ... > a comparatively weak relationship in many nonlinear (even completely ... mean that you are apt to have little sense from a correlation. ...
    (sci.stat.math)
  • Rank order
    ... I have a dataset with 20 years and in each year 40 countries are ranked ... I need to check if the ranking are statistically the ... that the correlation between ranking is 0. ... Sorry if this is a silly question, but I would really appreciate some ...
    (sci.stat.edu)
  • Re: US refused to admit they lost the Olympic to China
    ... medals and China winning the most golds, ... "Every four years, the question of how to rank countries comes up," ... U.S. on top while others show China," Wittkowski added. ... The new method of ranking is unique in a way that it does not assign ...
    (soc.culture.china)
Loading