"R" == Rich  <someone@somewhere.com> writes:

R> In infinite wisdom Joseph Lazio answered:

R> Statistics if properly done are descriptive of group distributions,
R> they are not predictive.

This seems a bit extreme.

R> The simple truth is extreme? How so?

I can use the known statistics of a group to predict the likely
properties of a new member that is discovered.

R> You can 'predict' that the new member is average. [...]

R> The only cases where you can make reasonably accurate predictions
R> are when the thing being predicted is not random. You can
R> reasonably predict that the next NAACP president will be black, and
R> indeed this probably applies to the membership. But this is not
R> much of a prediction.

I think your choice of examples illustrates a crucial difference
(between physical and social sciences).

Pick a star at random in the Milky Way.  I predict that that star is
likely to be an M dwarf.  Indeed, if you'd like me to be more
specific, I can quote the odds that the star will be of any given
spectral type.

Pick a pulsar at random in the Milky Way.  I predict that its velocity
is likely to be around 500 km/s.  More specifically, the odds are
roughly 95% that its velocity will be between 0 and 1100 km/s.

Similarly, as Louis says, finding more members of a group improves
my estimates of the group statistics.

R> Estimates? It does indeed give you more information about the
R> group.  But the fact remains, stats are descriptive, not
R> predictive. Will more information allow you to predict the mass of
R> the next planet discovered (...)?  No, it will not.

This sounds to me to be a semantic difference.  Given the current mass
distribution, the next planet to be discovered via the radial Doppler
technique is more likely to have a mass less than that of Jupiter.  

-- 
Lt. Lazio, HTML police   | e-mail: jlazio@patriot.net
No means no, stop rape.  | http://patriot.net/%7Ejlazio/
sci.astro FAQ at http://sciastro.astronomy.net/sci.astro.html