Double negatives, false positives, Type II errors... Double-sigh (;-)

>At that point the FLARM becomes completely useless to everybody
... and that puts us back to exactly where we are now!

> We need that data to evaluate it.  The data isn't just one of those
things that's nice to have, it's absolutely essential to determine
whether FLARM is an inferior offering to what CASA has in mind for
us.
...no we don't, 'cause then the statistically pedantic amongst us would
observe that ADS-B failure rates for GA/Commercial are not likely to be good
indicators of FLARM failure rates in our sport, and we would be no further
down the path to continuous improvement than we are now.

I'd rather pay a grand now for a commercially-manufactured FLARM which will
have marginal but increasingly important benefits as take-up increases, and
trust that the commercially astute manufacturer will give me software
upgrades when experience allows us to tweak the algorithms for even better
reliability.  Seems to me that so long as it's a good direction to be going
in, we should move in that direction, even if we know there will be further
improvements along the way (oops, isn't that just about a definition for
continuous improvement?).

LOL

Brett Kettle 

-----Original Message-----
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] On Behalf Of Mark Newton
Sent: Tuesday, 5 July 2005 11:47 AM
To: Discussion of issues relating to Soaring in Australia.
Subject: Re: [Aus-soaring] FLARM


On 05/07/2005, at 11:03 AM, Geoff Kidd wrote:
>
> Double negatives aside ............ how do you rate the double  
> positive of a warning that DOES occur when you have no idea that  
> the other ship is aiming for you HEAD-ON.



Sigh.

There are four possible modes that a system like this can be working in.

  * True Negative (Unit is silent, no threats nearby)
    This is where the typical FLARM installation will spend almost
    all of its life;  the proverbial NULL state.  As long as you have
    some way of testing whether it's still working, this is where you
    WANT it to spend its entire life.

  * True Positive (Unit is alerting, threat is actually nearby)
    This is where the FLARM is correctly alerting you to a potential
    midair.

  * False positive (Unit is alerting, no threats are actually nearby).
    If FLARM units do this more than a handful of times, they will
    produce an extremely powerful psychological reaction in the pilot,
    who will say, "The unit is dodgy, it doesn't work, I'm turning it
    off."  At that point the FLARM becomes completely useless to
    everybody.

  * False negative (Unit silent, actual threat nearby)
    This is at least as dangerous as having no FLARM unit at all,
    possibly more so if the pilot has formed the idea that he can
    compromise his lookout if the FLARM is going to tell him about
    midairs anyway.

For a FLARM to be 100% reliable, there would need to be 0 incidences
of the last two categories.

Currently we can probably guess that the false positive category
is pretty low, because anecdotal evidence about the usefulness of
FLARM seems positive, and if the people telling the anecdotes had
been pissed off by errors and turned it off they wouldn't be raving
about its virtues.

But we presently have no data whatsoever about the false negative
category.

Would FLARM be worth installing if the possibility of a false negative
was greater than the possibility of a true positive?  I'd think not.

Mike B reckons the ATSB is comfortable with ADS-B having a 25% false
negative rate (because it's only expected to prevent 75% of midairs).
Do we have any data whatsoever on whether FLARM's false negative
rate exceeds that of ADS-B ?

We need that data to evaluate it.  The data isn't just one of those
things that's nice to have, it's absolutely essential to determine
whether FLARM is an inferior offering to what CASA has in mind for
us.

   - mark

--------------------------------------------------------------------
I tried an internal modem,                    [EMAIL PROTECTED]
      but it hurt when I walked.                          Mark Newton
----- Voice: +61-4-1620-2223 ------------- Fax: +61-8-82231777 -----


_______________________________________________
Aus-soaring mailing list
[email protected]
To check or change subscription details, visit:
http://lists.internode.on.net/mailman/listinfo/aus-soaring

_______________________________________________
Aus-soaring mailing list
[email protected]
To check or change subscription details, visit:
http://lists.internode.on.net/mailman/listinfo/aus-soaring

Reply via email to