NavList:

   

Reply

Alex,

On Oct 25, 2004, at 11:33 PM, Alexandre Eremenko wrote:

>> It looks like you made the right choice
>> getting that snowy sextant rather than a Husun.
>
> I also like it. Still have to try really large distances.
> (It allows to 140 degrees!)
> I am afraid I will need another place than my balcony for this:-)
>

The US Navy Mark-series sextants, of WWII vintage and up to the 60s or
so, also read to 140 degrees or so.  I always thought that one reason
for this might have been using the instrument for rangefinding along
the coast; another plausible reason would be altitude shots with an
artificial horizon in the ice, as George Huxtable I believe pointed
out.  There are a few U.S. Navy navigators who frequent our list now
and again; perhaps one of them could tell us.

> But do you agree that that bad observation HAD to be rejected?
> It was evident to me immediately when I read my sextant.
> My reason and all what I know about statistics imply this.

I used to do this with data, but don't do it as much anymore, perhaps
because I'm getting older and less caring.  I think I've also noticed
over the years that I don't improve the precision very much; the
answers are much the same, both the numbers and the science.  Also one
begins to wonder about the repeatability of one's experiment.  This is
mostly with agricultural statistics, where the coefficient of variation
is close to 10% if you're lucky.

One most certainly would discard a blunder, such as reading the wrong
degree off the arc.  This error was borderline.  It was out beyond two
standard deviations, but just barely.  I think if the error is
detectable immediately there is more justification in rejecting it than
later during data reduction.

I think when one is trying to refine one's technique, examining all the
outliers and questioning their provenance is a vital exercise.  Also,
fitting data to some model can be an iterative process, where the
technique, data interpretation and model are altered to provide a
better fit.  [I read an example one time of Newton fooling with tide
data to fit a lunar model, and the person telling the story was trying
to make the case that Newton was cheating, fudging data.  I don't think
this can be the case during model development.  During model
validation, that's another story.]

Sorry for the non-answer to your question.

Fred

   

Reply