View Single Post
  #7  
Old December 14th 16, 08:44 AM posted to sci.astro
Poutnik[_5_]
external usenet poster
 
Posts: 68
Default 1) the negative paraxes...

Dne 13/12/2016 v 09:44 Martin Brown napsal(a):
On 12/12/2016 06:42, Poutnik wrote:



For exactly the reason I stated in the part of my reply that you
snipped: Negative values are unphysical, but form the part of the
statistical distribution of values that happen to lie below zero when
the mean is close to zero.


Positive and negative noise values are equally unphysical.


But you only know for certain that the negative values are unphysical
the positive ones could be real to within some measurement error. Later
more refined experiments may be able to narrow down the error bars.


Later experiment can. But I speak in context of this one.
These small values are not statistically justified,
as there is high probability it is just a noice.



Imagine that someone plotted a graph of, say, a spectrum (with low S/N),
and wherever the plotted flux was below zero, they simply truncated it.
Would you be happy with that? I wouldn't.


IMHO He should truncated all measurements
with zero belonging to CI of the measurement (mean) value,
as with statistically insignificant difference to zero.


No. Provided that it is stated somewhere what the limits of detection
for the method actually is then the value determined even if it is
negative is more useful to later researchers than a "below LOD" flag.

One thing is raw data, other thing is published processed data.
The limits should be available to a team of original data.

Such a limit can be estimated from the fluctuation around zero,
for stars where expected value is low enough.

--
Poutnik ( The Pilgrim, Der Wanderer )

A wise man guards words he says,
as they say about him more,
than he says about the subject.