NavList:

   

Reply

On Oct 26, 2004, at 6:06 PM, Alexandre Eremenko wrote:

> So we generally agree on philosophy of blunders.
> Now let us return to the particular lunar distance
> we were discussing.
> It was immediately clear to me that this was a blunder.
> So I rejected it.
> Then, on request of some list members, I reduced this
> erroneus observation.
> The reduction only confirms that this was a blunder.
> So I rejected it correctly.
>
> And now, to my great surprise, some people question my procedure.
>
> Alex.
>
>

Alex,

I don't think anybody has questioned your procedure.  I agree that if
an observation is immediately obvious as a blunder it should be
rejected.  But if you reject observations often, especially more than
one per session, I would say this is not a good idea or an indication
that your technique needs work.  After reduction, the observation was
only just barely detectable as a blunder by objective test, being 2
standard deviations away from the mean.  So it was a borderline case,
but one where conventional wisdom says rejection is OK.  If it had been
one standard deviation away, you probably wouldn't have rejected it,
while there would be little doubt if it were three standard deviations
away.  I think problems arise when one starts rejecting observations
that probably should have been kept, especially after data collection
is complete.

Fred

   

Reply