NavList:
A Community Devoted to the Preservation and Practice of Celestial Navigation and Other Methods of Traditional Wayfinding
Re: Accuracy of sextant observations at sea
From: Mike Burkes
Date: 2010 Nov 25, 18:55 -0800
From: Mike Burkes
Date: 2010 Nov 25, 18:55 -0800
Hi folks, graphing the obs Hs vs time and a line of best fit compared to the slope should reveal the "outliers" or "fliers" as we say. I believe Peter's "plotting of the 5 succesive shots" is saying the same thing.
Mike Burkes
626-833-1521
Date: Fri, 26 Nov 2010 07:16:41 +1100
Subject: [NavList] Re: Accuracy of sextant observations at sea
From: piterr11@gmail.com
To: NavList@fer3.com
Mike Burkes
626-833-1521
Date: Fri, 26 Nov 2010 07:16:41 +1100
Subject: [NavList] Re: Accuracy of sextant observations at sea
From: piterr11@gmail.com
To: NavList@fer3.com
Antoine Couette wrote:
The "weakest point" of Celnav is certainly the quality of your horizon + the uncertainties of your actual DIP correction.
I'd say the "weakest point" is weather precluding observations at all. But that's just a minor quibble compared to below, where you say:
"I also have taken the habit - some 30 years ago and whenever achievable since that time - to take 5 successive shots (spaced about 1 minute apart) and to use their average value both in times and CORRECTED heights. AVERAGING definitely narrows errors down."
If those "shots" include significant outliers (really wrong) then averaging can DECREASE the accuracy - you can end up with a result worse than adopting any of the other sights, with the obvious exception of the really off one.
Averaging is a good technique for decreasing random error given enough data points AND NO OTHER WAY OF DETERMINING WHAT IS GOOD OR BAD.
That is simply not the case here. The apparent rise or fall of the celestial body being observed is known (=easily calculated or estimated). Therefore those " 5 successive shots" can be plotted and compared with the fact of that slope. Any significant outliers can be discarded (either its wrong or all the others are) and you are left with a picture of your sights to contrast with the slope. If you then favour each of them equally then the methodology approximates averaging, but unlike a number-crunching exercise you can get to exercise your own judgement, perhaps based on your memory of which sights seemed at the time to be good ones, or the contrary.
Comparing multiple sights against slope "definitely narrows errors down". Averaging is not the right tool for this job. Too crude.
"I also have taken the habit - some 30 years ago and whenever achievable since that time - to take 5 successive shots (spaced about 1 minute apart) and to use their average value both in times and CORRECTED heights. AVERAGING definitely narrows errors down."
If those "shots" include significant outliers (really wrong) then averaging can DECREASE the accuracy - you can end up with a result worse than adopting any of the other sights, with the obvious exception of the really off one.
Averaging is a good technique for decreasing random error given enough data points AND NO OTHER WAY OF DETERMINING WHAT IS GOOD OR BAD.
That is simply not the case here. The apparent rise or fall of the celestial body being observed is known (=easily calculated or estimated). Therefore those " 5 successive shots" can be plotted and compared with the fact of that slope. Any significant outliers can be discarded (either its wrong or all the others are) and you are left with a picture of your sights to contrast with the slope. If you then favour each of them equally then the methodology approximates averaging, but unlike a number-crunching exercise you can get to exercise your own judgement, perhaps based on your memory of which sights seemed at the time to be good ones, or the contrary.
Comparing multiple sights against slope "definitely narrows errors down". Averaging is not the right tool for this job. Too crude.