Richard Ulrich wrote:
> On Wed, 7 May 2008 14:59:36 -0700 (PDT), Alex Lee
[..]
>> Is there a way that I can find (mine) two groups in a given
>> population.
[..]
> Sort them in order. Draw a line.
Ah, so simple. But where exactly?
For example, why draw a line between sorted third and fourth and not 2nd
and 3rd?
> If you want, you can evaluate a statistic that measures
> the distance achieved between groups.
Maybe we should try to minimize something ?
What? Sum of (inter-cluster) variances ?
What if the data are: 500ml, 700ml, 1L, 1.2L, 1.8L, 2.1L ?
This looks more like three groups, not two.
Alex,
If we know that there are exactly TWO groups, why don't we know who
belongs to which group? Lost the labels? The example is somewhat
unconvincing, you can surely distinguish men and women using other means
with no need to check their drinking habits.
JoJO
|
