NavList:
A Community Devoted to the Preservation and Practice of Celestial Navigation and Other Methods of Traditional Wayfinding
Re: Rejecting outliers
From: Peter Hakel
Date: 2011 Jan 1, 13:33 -0800
From: Peter Hakel
Date: 2011 Jan 1, 13:33 -0800
From: Fred Hebard
Peter,
Those are simulated data, not real data.
Fred
=============================================================================
Fred,
Would you be comfortable with a procedure that fails in the case of a simple (and yes, simulated) data set, for which you know a priori what the right answer is? Would you trust your new calculator to process real data if it gave you 1+1=3 in an "artificial" test the first time you use it?
The success with examples that I provided are necessary for the method to be valid; I never said it was sufficient. You seem to imply that there is no value in working with simulated data. Quite the opposite is true; you have to be able to get 1+1=2 (or, in this case, computed altitude #5 close to 14) before you can even hope to deal with real data with any confidence.
For example, before I successfully processed real-life meridian transit data (long ago) and the recent ex-meridian data set, I put the parabolic-fitting procedure through the simplest simulated test of having three non-colinear points (image attached).
If real data becomes available this time around, once again I'll be happy to work with it and make adjustments in the algorithm, if necessary.
Peter Hakel
Peter,
Those are simulated data, not real data.
Fred
=============================================================================
Fred,
Would you be comfortable with a procedure that fails in the case of a simple (and yes, simulated) data set, for which you know a priori what the right answer is? Would you trust your new calculator to process real data if it gave you 1+1=3 in an "artificial" test the first time you use it?
The success with examples that I provided are necessary for the method to be valid; I never said it was sufficient. You seem to imply that there is no value in working with simulated data. Quite the opposite is true; you have to be able to get 1+1=2 (or, in this case, computed altitude #5 close to 14) before you can even hope to deal with real data with any confidence.
For example, before I successfully processed real-life meridian transit data (long ago) and the recent ex-meridian data set, I put the parabolic-fitting procedure through the simplest simulated test of having three non-colinear points (image attached).
If real data becomes available this time around, once again I'll be happy to work with it and make adjustments in the algorithm, if necessary.
Peter Hakel