A Community Devoted to the Preservation and Practice of Celestial Navigation and Other Methods of Traditional Wayfinding
From: Robert VanderPol II
Date: 2016 Jul 25, 22:17 -0700
I went back and looked at all the stats you've given over the last 6 weeks or so trying to tease out the best compromise between accuracy and usability.
The givens seem to be:
Max Dec 75o
Max Lat 70o
Decimal point from 54o
Errors that would be reasonable to me were obtained by using mostly No-Interpolation, some Interpolation and a bit of Sadler. The stats I found were:
RMS Alt Error >0.5' Max Error Remarks Posted
t<82o or t>98o ? 4% 2.9' No interpolation 15-Jun
82<t<87 or 93<t<98 Interpolation
87<t<93 0.4 12% 8.5' Sadler 18-Jul
In general :
About 0.2% azimuths greater than 0.5o in error with Max Az error ~ 2o
The value I would think most important would be max error. Yes, one would like to keep RMS and %>0.5’ as low as possible, but the RMS errors seem to be similar to or better than what one can practically expect from on the water CelNav. Given that I would be most interested in how badly I can get burned no matter how unlikely and use this as a basis for making decisions about how much to trust my results when making navigation decisions.
It seems to me that the way to pick where one converts from using Interpolation to using Sadler is to calc the stats for 1o intervals and compare when the one method becomes less advantageous than the other.
So for the interval 86<t<87 what are the stats for both Interpolation and Sadler? And 85<t<86? And 87<t<88?
Or focusing just on the Max Error value that I prefer what are the Max Errors for Sadler and for Interpolation for each degree interval from 80o-90o?
With that I could create a graph and pick the point where they cross.
For deciding when to switch from NoInterpolation to Interpolation I would expect the interpolation to always give a better result (sometimes much better, sometimes slightly better) that is dependant on the situation one is in. Offshore no nearby hazards, don’t bother to interpolate. Inshore you go with Interpolation. The decision of when to change methods strikes me as somewhat dependant on the user’s risk tolerance so a comparative graph would allow a user to make their own educated decision. To that end what are the Max Errors for Interpolated and Non-Interpolated 0o-80o for 1o intervals?
Am I barking up the wrong tree focusing on Max Error?
Am I asking for more analysis than you have time for?