On 2/16/2016 8:48 PM, Steve Willner wrote:
In article ,
Jos Bergervoet writes:
Why are the 2 predicted curve in this picture slightly different?
https://www.ligo.caltech.edu/image/ligo20160211a

As the caption notes, one of the curves has to be inverted and
shifted by 7 ms because of the different detector locations and
orientations.

The data are shifted, not the predicted curves, those are
inverted but not shifted as can clearly be seen in the
ring-down region at the end.

The rest of the differences are presumably random
noise.

There will be noise in the data, but why in the predictions?

The signal to noise is only 5 sigma, after all. Figure 1 of
the published paper is clearer.

OK, we look at:
https://dcc.ligo.org/public/0122/P150914/014/LIGO-P150914_Detection_of_GW150914.pdf

There in Fig. 1, the predicted curves are called "Numerical
relativity" and they *still* have differences in shape, that
are not inversion or shift transformations! It is becoming
more and more intriguing.

Possible explanations:
1) For some reason (to make the curve look more "natural"?)
someone decided to add random noise to the computed results.
And they added *different* noise for Hanford and Livingston.
To me this seems a silly eplanation.
2) The results are different polarization componentsa (after
all you only need a 45 degree tilt to see the independent
other polarization for a spin-2 field.)
3) The numerical routines generate some numerical errors
visible as small random looking "ripples" in the computed
result. This seems likely since complex curved space-time
will enforce a complicated non-uniform grid in the 4
coordinates.

A combination of 2) and 3) seems most likely to me (I didn't
find any proof for it in the text, but I may have overlooked
it in the extensive list of papers that have accumulated.)

--
Jos