|
|
|
|
|||||
|
||||||
From: Bill (no email)
Date: Sun Oct 24 2004 - 01:32:22 EDT
> The problem with this approach is all values are given equal weight. The
> resulting slope is influenced by outliers (extreme values). In any case, there
> doesnąt seem to be any point in averaging once the slope is known. Think of a
> few data points that approximate the right slope. What is not wanted is
> anything that takes the chosen value away from there, as averaging might,
> depending on the data points that donąt follow the slope.
There appear to be workarounds to minimize the affect of outliers for the
mathematically inclined. I am a sailor first, attempting to learn more
about cel nav--past plug and chug. I practice both on land and water, and
need pencil-and-paper solutions I can execute on the water.
From that vantage point, I wholeheartedly agree as to calculating a slope
and fitting data points to it. Plot the observations. Calculate the slope
for that time period. Use that slope to determine your outliers. For those
of us who are not mathematicians, surveyors, engineers, historians,
astronomers or some combination thereof, I alluded weeks ago to the rational
and methods for calculating the slope (which were included in older versions
of Bowditch), as well as instructions for the computer or reduction tables.
Drill down in David Burch's site. If you missed it or have more interest as
of late, the staring point is:
http://www.starpath.com/index.htm
Celestial Navigation Section
David Plastic Sextants
How to take plastic sextant sights
Fit slope method
The destination URL as shown in my (Mac/IE) browser is:
http://www.starpath.com/online/celestial/sight_average.pdf
Bill
|