Southworth Det Eclips Binary Catalog

I found a very good catalog of detached eclipsing binary stars with
mass determinations "accurate to 2%". It is an ongoing catalog with
new systems being added as they are published.

http://www.astro.keele.ac.uk/jkt/debcat/

This would appear to offer a good preliminary sample with which to
test my hypothesis that the total masses of binary star systems (and
single white dwarfs) have distributions that are characterized by
preferred masses that are integer multiples of 0.145 solar mass.

Taking only 2012 and 2013 data from Southworth's catalog (I am only
interested in new mass determinations), I find a sample of 36 systems
with sufficiently narrow error bars for an adequate preliminary test
of the hypothesis.

Of the 36 test systems, 77% are located at 0.04 solar mass from one
of the predicted preferred masses. Roughly 22% are located at =/ 0.04
solar mass.

The total masses for the EcB systems cluster around the predicted
masses. A histogram of the + and - deviations is centrally peaked at
the generic multiple value.

These results seem much better than the results for the small sample
of neutron star binary systems, with the previously noted
heterogeneity in error estimates.

I am of course wondering if I have something to write home about yet,
and I am hoping that the readers of SAR will have some constructive
criticism.

Perhaps I might also get some advice on the best way to analyze the
data so as to clearly demonstrate what the data say, and what they do
not say.

RLO
Fractal Cosmology/Discrete Scale Relativity

In article , "Robert L.
Oldershaw" writes:

This would appear to offer a good preliminary sample with which to
test my hypothesis that the total masses of binary star systems (and
single white dwarfs) have distributions that are characterized by
preferred masses that are integer multiples of 0.145 solar mass.

Get out the envelope so that we can write on the back.

Taking only 2012 and 2013 data from Southworth's catalog (I am only
interested in new mass determinations), I find a sample of 36 systems
with sufficiently narrow error bars for an adequate preliminary test
of the hypothesis.

It's not clear how you use the error bars, unless you mean that they are
appreciably smaller than 0.145 solar mass.

Of the 36 test systems, 77% are located at 0.04 solar mass from one
of the predicted preferred masses. Roughly 22% are located at =/ 0.04
solar mass.

1/3 of .145 is about 0.048, close enough to 0.04. Let | be one of your
integral multiples and - denote 1/3 of this, roughly your 0.04. If the
masses are distributed uniformly, then you would expect 2/3 of them to
be located within 1/3 of the distance between two points. In other
words, these are those marked with ! below and the 1/3 are marked with
dots.

!.! !.! !.! !.! !.! !.! !.!
---|---|---|---|---|---|---

So your numbers indicate that the masses are consistent with a uniform
distribution.

On Wednesday, November 27, 2013 3:25:10 PM UTC-5, Robert L. Oldershaw wrote:
I found a very good catalog of detached eclipsing binary stars with
mass determinations "accurate to 2%". It is an ongoing catalog with
new systems being added as they are published.
.....
Perhaps I might also get some advice on the best way to analyze the
data so as to clearly demonstrate what the data say, and what they do
not say.

I took a look at the M1 and M2 values. I used only the values which
had error bars smaller than 20% of 0.145, which yielded 112 and 120
samples (M1 and M2 respectively).

I made a histogram of M1 and M2, modulo 0.145 Msun, with 20 bins in
each. If the mass values were clustered at multiples of 0.145 Msun,
then one would expected a peaked distribution at either end of the
histogram, and a trough in the middle of the histogram. In actuality,
there are no significant peaks or troughs. The distributions are
approximately flat to within the (Poisson) errors, indicating no
clustering of masses.

This can be quantified. I presupposed a hypothesis that there is no
clustering of mass values, i.e. flat distribution, and I can test that
hypothesis using a chi-square statistical test. I found chi-square
values of 23.4 (M1) and 20.5 (M2), for 20 degrees of freedom.

Given these values, the no-clustering hypothesis cannot be rejected
(significance values 23% & 43%). I.e. there is no support in this data
for clustered masses of either M1 or M2 at multiples of 0.145.

There are perhaps better ways to quantify this. I used a frequentist
approach, which is quick and dirty.

CM

On Thursday, November 28, 2013 4:08:37 AM UTC-5, wrote:
There are perhaps better ways to quantify this. I used a frequentist
approach, which is quick and dirty.

---------------------------------------------------

Just a quick comment before I get back to Thanksgiving activities.

If you have samples of over 100 systems then you are NOT using only 2012 and 2013 data.

I do not trust the 1990s Anderson data or even the newer Torres et al data.

[Mod. note: because they disagree with you, or for some objective
reason? -- mjh]

I am ONLY interested in 2012, 2013 and future analyses. This makes the
progress in testing the hypothesis much slower, but I think much more
trustworthy. I do not want to discuss this decision of mine, which is
based on studying the empirical data for over 30 years. It is my
choice and I think it is a very wise one.

I will read your post more carefully when time permits.

Happy Thanksgiving!

RLO
Fractal Cosmology

[Mod. note: reformatted and entire quoted article trimmed -- mjh]

On Thursday, November 28, 2013 4:08:37 AM UTC-5, wrote:
On Wednesday, November 27, 2013 3:25:10 PM UTC-5, Robert L. Oldershaw wrote:


Given these values, the no-clustering hypothesis cannot be rejected

(significance values 23% & 43%). I.e. there is no support in this data

for clustered masses of either M1 or M2 at multiples of 0.145.



There are perhaps better ways to quantify this. I used a frequentist

approach, which is quick and dirty.

----------------------------------------------------

WHOA!!!

I just noticed that you are testing M1 and M2 SEPERATELY!!!

My hypothesis is that it is the TOTAL SYSTEM MASS that shows
indications of preferred masses that are integer multiples of 0.145
solar mass.

I would be delighted if you would check M1+M2 values for ONLY 2012 and
2013 systems. Then we would be on the same page (in the same book?).

Best,
RLO

On Thursday, November 28, 2013 3:06:17 PM UTC-5, Robert L. Oldershaw wrote:
I just noticed that you are testing M1 and M2 SEPERATELY!!!

My hypothesis is that it is the TOTAL SYSTEM MASS that shows
indications of preferred masses that are integer multiples of 0.145
solar mass.

I would be delighted if you would check M1+M2 values for ONLY 2012 and
2013 systems. Then we would be on the same page (in the same book?).

I used the full table for M1+M2 (with error bars 0.2*0.145) yielding
a sample of 110. I also binned (M1+M2) modulo 0.145 into 20 histogram
bins. Again, no evidence of any clustering near 0.145 Msun (chi square
value of 25 for 20 degrees of freedom, significance level of 20% for
null hypothesis).

I don't really have the time to apply your arbitrary selection
criteria (systems measured years 2012 & 2013 only). I'm not aware of
any magical change that allowed researchers to suddenly determine
masses more reliably or accurately starting in 2012. I chose criteria
I could apply quickly and objectively. The error bars need to be
significantly less than your chosen multiple of 0.145 Msun, which is
why I chose 20%.

You are welcome to apply other more stringent selection criteria, but
beware you would be getting into the regime of very low counts where
significance would need to be assessed carefully (as well as number of
trials perhaps).

CM

On Thursday, November 28, 2013 4:07:30 AM UTC-5, Phillip Helbig---undress to reply wrote:


I answered this once, but the post did not appear. Here is a brief summary.

If the distribution was "uniform" for the +/- deviations away from a
predicted value, then the 0.01-sized bins between -0.07 solar mass to
+0.07 solar mass, with 0.00 as a predicted multiple of 0.145 solar
mass, would contain statistically equal numbers of "hits" - i.e., a
flat distribution.

This is NOT observed.

The distribution of +/- deviations is CENTRALLY-PEAKED. So it is your
hypothesis of a uniform distribution that is at odds with the data.

[Mod. note: why not show us a figure? -- mjh]

In article , "Robert L.
Oldershaw" writes:

If the distribution was "uniform" for the +/- deviations away from a
predicted value, then the 0.01-sized bins between -0.07 solar mass to
+0.07 solar mass, with 0.00 as a predicted multiple of 0.145 solar
mass, would contain statistically equal numbers of "hits" - i.e., a
flat distribution.

This is NOT observed.

First, the expectation is not flat, but Poisson.

The distribution of +/- deviations is CENTRALLY-PEAKED. So it is your
hypothesis of a uniform distribution that is at odds with the data.

Second: Other posters and I have shown, based on your numbers, that
there is no evidence for the fact that the UNDERLYING distribution is
flat (the OBSERVED distribution will not be flat, due to small-number
statistics) and there is no evidence for your peaks at integer multiples
of 0.0145 solar masses.

[Mod. note: why not show us a figure? -- mjh]

Third: Indeed.

The fact is that no-one except you sees any evidence at all for these
peaks. Even you see them only by cherry-picking the data (which you are
"not willing to discuss") and by changing the criteria with time (now it
is the total mass; it used to be the individual masses). Put all of
this together and the result is that you have convinced no-one. In
fact, what has probably happened is that your unsupported claims have
made people even more sceptical of DSR.

On Friday, November 29, 2013 2:45:33 AM UTC-5, wrote:
On Thursday, November 28, 2013 3:06:17 PM UTC-5, Robert L. Oldershaw wrote:

------------------------------------------------------

I don't know how many times I have to say it.

I am justified in defining a hypothesis and identifying a proper test
of it, as long as I am not engaging in some sort of biased
manipulation of the data.

Choosing only to allow data published from 2012 and into the
open-ended future seems scientifically to valid to me.

I have good scientific reasons to make this choice but it is a
separate discussion and I do not want to get bogged down in this issue
BEFORE we establish what the hypothesis and test I have chosen tells
us. This would just be avoiding the crux of the matter.

If you do not want to put in the effort required to analyze my more
constrained sample, fine. Just don't tell me you know the
correctness/incorrectness of the hypothesis before you do a fair test
with the BEST data available, and excluding decades-old data that is
suspect. I an not saying it is proven wrong; I am saying it cannot be
trusted for such an important hypothesis.

On Friday, November 29, 2013 2:46:35 AM UTC-5, Robert L. Oldershaw wrote:
On Thursday, November 28, 2013 4:07:30 AM UTC-5, Phillip Helbig---undress to reply wrote:


[Mod. note: why not show us a figure? -- mjh]

-----------------------------------------------------

Martin, I can assure you that I have plans to write up and publish a
substantial paper on preferred stellar masses that includes the
eclipsing binary data, the white dwarf data, the planetary nebula
nuclei data, the exoplanet system data, and the neutron star data.

But don't hold your breath. I am one modestly skilled person working
in a less than encouraging environment. This thread is a test of how
much opposition I can expect in doing research on this very radical
hypothesis. I am also wondering if anyone will say: "That looks
interesting and worthy of being pursued further". I would ideally like
to get those with more advanced skills in statistics and data
presentation to participate in this research. That may be an
unreasonable dream, but I have endless hope, plenty of time and the
perseverance of a mule.