Re: FarmingFisicistsFumble

Thus spake Oz <Oz@xxxxxxxxxxxxxxxxxxxx>
>
>I am a tad bothered that:

I would also have liked this to be a nicer fit, but console myself with
the thought that it nowhere near bad enough for rejection. It is
compensated for by the Astier fit, with better data.

The problems are in the Riess data. We have standard fit chi^2=174,
teleconnection fit chi^2=183. That corresponds to rejection at 80%
confidence for the standard model, and at 85% for the teleconnection.
But the fact that Astier selects only some of the low z data and
processes it more accurately makes me suspect that the true reason is
that Riess's data came from a number of different sources and was not as
uniformly prepared.

By a rough estimate I had it that the difference in chi^2 for the Riess
data is almost entirely due to the five highest data points. Five points
is not generally enough for statistical significance.

>1) Fig 2 shows SN 1997ff as well off your line. A comment below the
>figure would be good.

Its less than 2sigma, and there is this question mark over gravitational
lensing. If the magnification has been over stated then the plotted
point will drop below the line. In fact, lower down the plot quite a few
supernova are just as far below both curves. I believe all points here
are corrected for gravitational lensing, but as I have no knowledge of
what the plot looked like without it is difficult to get a feel for what
might be going on.
>
>2) Deviations on both halves for fig 3 are Not Good.

A remarkable number of points are outside 1sigma errors. I don't know if
I can get Bob to run his program on just the Hubble points, but I bet
chi^2 is well above the number of df (degrees freedom, meaning the
number of points plotted). It also looks to me like the Hubble data is
consistently low (do you agree).

>Can you modify your formula to get a better fit?
>
Not much. These curves are plotted at Omega=1. Larger values of Omega
bring the curve downwards towards the data points, but fig 5 at Omega
=1.15 still shows the curve above the point. Omega would be still larger
with no cosmological constant, but then I get very uncomfortable about
largish values of Omega_R.



Regards

--
Charles Francis
Please reply by name
.

Relevant Pages

  • Re: Curve Fitting
    ... results from many nonlinear fits. ... not really fit the shape of most common "s" surves. ... These curves ... its pchip module. ...
    (comp.soft-sys.matlab)
  • Re: Curve Fitting
    ... results from many nonlinear fits. ... not really fit the shape of most common "s" surves. ... These curves ... but then have soft tails. ...
    (comp.soft-sys.matlab)
  • Re: This is What Einstein Actually Did.
    ... You have seen the curves. ... Fit the 1920 Angstrom curve ... therefore will exhibit bigger magnitude changes at the same observer distance. ... Henri. ...
    (sci.physics.relativity)
  • statistical comparison of exponential association fits
    ... the snippet below simulates 4 different two-parameter exponential association curves ... the simulations and results are displayed; all other resulting statistical data are saved... ... % simulation parameters ... whether two curves differ using the outputs of FIT ...
    (comp.soft-sys.matlab)
  • Re: Astronomical Software
    ... >> the same results as Astier or Riess, and I think it is because you are ... model can be rejected just on the ground of accuracy of fit. ... H0 will lead to a greater error in Omega. ...
    (sci.astro.research)
Loading