An unmodelled component in spectrographic measurements of local stars

[snipped sci.physics.research]

In article ,
Oh No writes:
I have redone the test for an unmodelled component in spectrographic
measurements of local stars
.....
http://arxiv.org/abs/0903.5193

This is an interesting idea, but the statistical treatment seems very
odd. I agree non-parametric statistics are appropriate, but why not
apply them directly to determining the slope (or more appropriately
the curvature) of the velocity distributions of large numbers of
stars rather than deriving a binary yes/no for many subsets? And why
choose subsets in such an arbitrary way rather than on an
astrophysical basis?

Leaving aside the question of methods, if you arbitrarily adjusted
the Hipparcos distance scale to make the effect go away, how much
would you have to adjust by?

It will be about seven years before Gaia gives direct measurements of
the radial velocity of individual stars.

Maser measurements might provide results sooner than that. At least
one source ( http://arxiv.org/abs/0902.3913 ) seems inconsistent with
your results, but I haven't done a detailed comparison.

--
Steve Willner Phone 617-495-7123
Cambridge, MA 02138 USA

Thus spake Steve Willner
[snipped sci.physics.research]

In article ,
Oh No writes:
I have redone the test for an unmodelled component in spectrographic
measurements of local stars
....
http://arxiv.org/abs/0903.5193

This is an interesting idea, but the statistical treatment seems very
odd. I agree non-parametric statistics are appropriate, but why not
apply them directly to determining the slope (or more appropriately
the curvature) of the velocity distributions of large numbers of
stars

You haven't followed exactly what I was doing. My calculation of the
slope came from http://arxiv.org/abs/0812.4032v2. Together with the
slope from CO and HI given by Combes (with which it agrees) this was
used in this paper only to motivate the analysis.

Measurements on stars are not extensive enough to calculate the curve
itself, which is why I have only looked at the slope. I did try
calculating directly from the velocity distributions of stars, but found
too much variability in results depending on colour to feel I was
getting a meaningful result. According to my analysis in
http://arxiv.org/abs/0901.3503 the structure of the spiral arm is that
stars near pericentre are likely to be on the inside of the arm, nearer
the sun's position. One actually finds much steeper slopes for late
types stars and reversed slopes for early types - actually that's point
supporting my analysis which I had forgotten (I calculated slopes when I
was working with the original Hipparcos data, quite some time before
analysing spiral structure, . I shall redo the calculation using the
HNR).


rather than deriving a binary yes/no for many subsets?

The binary yes/no test is not a way to calculate the slope, but simply a
way to ascertain whether there is an unmodelled component.

And why
choose subsets in such an arbitrary way rather than on an
astrophysical basis?

Given the noisy nature of the distribution, as well as the structure
which exists within it, it took quite some time to come up with any
meaningful test. The division into quadrants negates problems due to
structure. The division into colour bins is vaguely astrophysical. I had
thought to subdivide into giants and dwarfs from the H-R diagram, but
there is a lot of extra work and I could see no benefit. The only
important thing for the test is to divide into disjoint subsets.


Leaving aside the question of methods, if you arbitrarily adjusted
the Hipparcos distance scale to make the effect go away, how much
would you have to adjust by?

I can't make the effect go away by simply adjusting the distance scale,
because the effect is seen in the V-W plane, but not along the U axis.
With a 10% increase in parallax distance, the effect in the V-W plane is
still significant at 97.6%, but this would cause one to reject results
on the U-axis. With a 15% increase in parallax distance there is no
indication of an effect in the V-W plane.

It will be about seven years before Gaia gives direct measurements of
the radial velocity of individual stars.

Maser measurements might provide results sooner than that.

I don't think so. By direct measurement I mean change of parallax
distance over a period of time, or rate of change of proper motion over
a period of time. I am not an expert on masers, but I believe they
depend on spectroscopy, which would not reveal the effect I am testing
for.

At least
one source ( http://arxiv.org/abs/0902.3913 ) seems inconsistent with
your results, but I haven't done a detailed comparison.

That paper only includes 18 sources - far too few for a statistical
analysis. To be honest I find it a bit worrying that the authors think
they can get results favouring a given number of arms, or that they can
do a meaningful calculation of the pitch angle of an arm with a sample
this small. In a statistical science I would think that the statistical
training of the practitioners needs to be better than that. The same is
true of papers by Georgelin & Georgelin, and by Russel who advocated
four arm in the first place. I think my analysis in http://arxiv.org/abs
/0901.3503, showing a two armed spiral with a much lower pitch angle
based on the observed HI distribution, on the velocity distribution of
20574 local stars, and on dynamical analysis with numerical solution is
more meaningful.

Regards

--
Charles Francis
moderator sci.physics.foundations.
charles (dot) e (dot) h (dot) francis (at) googlemail.com (remove spaces and
braces)

http://www.teleconnection.info/rqg/MainIndex