NavList:

Jeremy,

I ran your on-transit data through my rapid-fire fix program (see attached input data file) and obtained results that agree very well with those from my earlier posting (calculated by a least-squares parabolic fit).  This way I obtain:

1900L fix:
Latitude:     N    21 degrees  47.3'   :  standard deviation =   1.6'
Longitude:  E  130  degrees  04.5'  :  standard deviation = 44.6'

From looking at the azimuths (recorded to 0.1 degrees) the meridian crossing (viewed by the observer) occurred at UT=9:51:19.  The moon azimuth was changing at the rate of 0.1 degrees per 14 seconds, so this could serve as a rough error estimate, say +/- 7s.  Most likely this could be refined by taking more data points around LAM and do some kind of fitting and/or interpolation.

My standard deviation in longitude is much larger than Antoine's so his procedure must be performing a more sophisticated statistical analysis than my simple program does.  Perhaps I am overestimating it, considering that my mean longitude value is very close to Antoine's.  There may be a rigorous way of narrowing its variance down, instead of just using the standard formula. On the other hand, my "sigma" values are consistent with the notion that meridian transits are good for calculating latitude and less good for measuring longitude.  I remember Frank mentioning in the past the notion of the "error ellipse" where you would have a smaller error along the azimuth line and a larger one along the direction perpendicular to it.  Something to think about, perhaps other list members can help me better understand this point and resolve the apparent contradiction.

Also, while in this case I do not see any outliers in the data (all intercepts are TOWARD and generally increasing with time), there is a curious jump in value from 7.5 at 9:48:15, to 9.3 at 9:48:51.  I wonder what happened there, although it may be nothing important.


Peter Hakel