NavList:
A Community Devoted to the Preservation and Practice of Celestial Navigation and Other Methods of Traditional Wayfinding
Re: Systematic error and its resolution
From: George Huxtable
Date: 2007 Apr 6, 18:43 +0100
From: George Huxtable
Date: 2007 Apr 6, 18:43 +0100
Following Peter Fogg's contribution obout systematic error, let's consider that topic a bit more carefully. He rightly separates out the possibility of systematic error in timing, and its effects on longitudes and latitudes. But then, the second of his "two sources", is- | 2. A systematic error in altitude, such as using corrections with the | wrong sign, damage to the sextant, anomalous refraction, and the like. and he then continues- | Correction for systematic error | | If the error is systematic, each position line will be displaced equally by | the amount and in the direction of the error. Trouble is, that doesn't accord with his examples above, of what he calls a "systematic error in altitude". Take "using corrections with the wrong sign". Well, if you applied the refraction correction with the wrong sign, it wouldn't make much difference for high-altitude objects, but would systematically put lower bodies into the wrong place. Take "damage to the sextant". I wonder what damage he is thinking about here. If any such damage has affected the scale calibration, over different parts of the arc, or given rise to drum eccentricity or collimation error, or if a user has misinterpreted the sign of the "box corrections" on the certificate pasted inside his sextant box (or ignored it), these will all be systematic errors, but all will vary with altitude. Take "anomalous refraction". I wonder if he is thinking about "anomalous dip" here, which varies with the low-level refraction in the path from the horizon to the eye. Yes, that would indeed give rise to a systematic error, common to every observation. But NOT anomalous refraction, in the light path from the observed body. That can indeed have serious consequences, as is clear to anyone who has watched the distortions that often occur in the shape of the disc of a low Sun, and wondered how its altitude might be affected. And different altitudes will be affected by different amounts. Yes, indeed, there are causes that will give rise to a common, equal, error, in every corrected altitude, and getting the index error wrong is perhaps the most obvious of them. And only when it's known that the cause is such as to give rise to equal errors at all altitudes, will the method, that he described, apply. It applies, not to "systematic errors", in general, but to systematic common-altitude errors only. And there's another difficulty, that he doesn't go into, but it's a crucial one. We have recently gone at great length into the analysis of error-triangles in which the basic assumption was that all systematic errors had been corrected, and only random scatter remained. Peter is now dealing with a case where it's assumed that the errors are systematic ones (and he says so), and he must be presuming that there are no (or insignificantly small) random errors. If it was somehow known that the only errors in a round of observations were common systematic ones and that there was no random scatter, then I would concur that his construction method, to eliminate those errors, is a valid one. But how on Earth can anyone TELL that, from a single round of observations? In fact, both assumptions are unrealistic; there will always be an unknown mix of random scatter and systematic error, and how are they to be disentangled?. Only by considering the statistics of many such rounds of sights could you reach any such conclusions; not that I'm advocating that. Just by chance, if there was zero systematic error, one observation in four will show the displacements all in the same "systematic" direction, which would indicate to Peter that some such common error was indeed present, even if it wasn't, and he would "correct" the triangle accordingly. But that would have been done on the basis of a wrong assumption, so would he have "improved" anything at all? I doubt it. The technique Peter puts forward has been proposed a number of times, notably by George Bennett. I am going back to the example Bennett gave by distant memory, but as I remember he was teaching a group of students who were all measuring sextant altitudes of certain bodies. All had arrived at consistent results except one, and Bennett was later able to show, by a similar technique to Peter's, that there must have been a serious error in the index correction that had been used in that case, and to put it right. You can see that in that particular example, the method was indeed valid. The consistency of the other students' results showed clearly that random scatter was indeed small, by comparison with the major errors shown by the odd-one-out. How often does that apply in real-life? Peter's conclusions, such as that his "improved" fix will always lie inside the triangle if the azimuths spread by more than 180 degrees, and, indeed will be at its centre, are over simplistic, in that they apply only if it is known that there is no random scatter, or if the systematic error is sufficiently great as to completely overwhelm it. Indeed, if you were to find that in every case the three position lines in such a triangle were indeed displaced in that same "systematic" manner, in a series of many rounds of sights, that should excite suspicion that somewhere a systematic error exists, which needs to be tracked down and corrected. George. contact George Huxtable at george@huxtable.u-net.com or at +44 1865 820222 (from UK, 01865 820222) or at 1 Sandy Lane, Southmoor, Abingdon, Oxon OX13 5HX, UK. --~--~---------~--~----~------------~-------~--~----~ To post to this group, send email to NavList@fer3.com To , send email to NavList-@fer3.com -~----------~----~----~----~------~----~------~--~---