NavList:
A Community Devoted to the Preservation and Practice of Celestial Navigation and Other Methods of Traditional Wayfinding
Re: Real accuracy of the method of lunar distances
From: Fred Hebard
Date: 2003 Dec 31, 19:31 -0500
From: Fred Hebard
Date: 2003 Dec 31, 19:31 -0500
Jan, One minor point, correcting an earlier post of mine, is that the standard deviation is equal to the square root of (the sum of squared deviations divided by the degrees of freedom, not by the number of observations). In this case, the numerator would be divided by 33, not 34, including all observations in the data set. From your posting, I infer that the standard deviation was constructed from the deviation of the mean of the individual observations from the "actual" value as determined by chronometer. So that for Bolt's data there were 34 sets of observations with 34 deviations (ignoring for the present the subdivision into sun lunars, star lunars and two-sided lunars). The error for these observations should be the same as what would be obtained if each of the 34*6 individual observations was calculated, unless Bolt's sextant had a significant change in accuracy over the course of the voyage or unless his technique were improving or getting worse over the course of the voyage. You could get a feel for this by looking for a change in the deviations over the course of the voyage, perhaps by plotting them against time and perhaps looking for a significant difference from zero of the slope of a straight line fit to the data. I am saying that the errors within sets of observations would equal errors between sets, for this particular model. Yes, I was altering the standard deviation to the standard error of the mean. You appeared to be constructing confidence intervals based upon the t statistic. The test statistic in this case would be t = (deviation from chronometer time) / (standard deviation/square root of number of observations). For large numbers of observations, t is equal to 2 at 95% confidence, 2.8 at 99% and 3.6 at 99.9% confidence. I will take your word for it that a value of 3 gives 99.7% confidence. Given the above, it is my belief it would be appropriate to divide by the square root of the number of observations making up a single observation. So perhaps it's not quite as bleak as your analysis indicated. Fred On Dec 31, 2003, at 6:20 AM, Jan Kalivoda wrote: > No, Fred, > > Errors reported in my posting were given for averaged measurements > from six shots in every case. Each such average was taken as the > *single* observation and the error was compared with the corrected > chronometer time only for this average, not for each of six original > shots. Therefore, no error referred to in my posting can be divided by > the square root of six, as you propose (as if you search for the > standard error of the mean from the set of six measurements). > > > Yours, Jan Kalivoda >