NavList:

   

Reply

At 12:08 AM 3/31/02, Bruce Stark wrote:

>In a message dated 3/29/02 4:12:37 PM, george@HUXTABLE.U-NET.COM writes:
>
><< It's clear that this long-winded reiteration process does not lend itself
>to hand-calculation, though a computer would do it in a twinkle.
>
>If anyone (and here I am thinking particularly of Bruce Stark) can suggest
>a way around this problem, I would be pleased to hear about it. >>
>
>Bruce responds: Well, I had my first go at it this evening. Calculated the
>altitudes for that first set of Pollux distances taken Monday night, using
>GMT 04:00:00. That's three-quarters of an hour early of the truth. Pollux's
>altitude came out O.'4 different from the that gotten with the correct GMT.

....

>Bruce

I want to tell the list how interesting I am finding this effort to use a
time insensitive method to find the time, and specially how some actual
observations were given, and 'gardened' for out riders.

I will admit, I had the sincerest doubts about the wisdom of discarding two
out riders because they 'looked wrong'. I seem to remember that gardening
for outriders is not in itself suspect. But the procedure if I recall was
to pick out values on the tails of a distribution - and from both sides.....
  Anyway, I sang along by finding a least squares fit to ALL the
observations, and agreeing with George on the result of picking off the two
wild out riders in one dataset offered on the list.    So I should not
complain at what works.
I had a moment of weakness I will admit, to working on the kind of slope
estimator George had suggested as a function of altitude and latitude etc.,
but the weakness passed in time, thank goodness!  :-)  particularly as I
enjoy using the non linear regression package NLREG. It  is an obvious and
widely used tool for making a good shake at estimating weights for factors
affecting a variable in this way.

Sincerely

Brian W


Brian Whatcott
   Altus OK                      Eureka!

   

Reply