How to analyse the small number of MAC’s in Oz? Some have said the number is small enough to look at each accident individually. Yes that’s fine – but that does not negate the value of a statistical approach.
And the best statistical approach is start with the exponentially reducing weighting which I described. This applied to each GAAP aerodrome individually, to yield the best estimate of current MAC rates. (Later a model is fitted to those averages).
Others have said “poop happens” at random - and there is no point in analysing the MAC numbers.
And Peuce wrote in #281: “My point is that if we are to compare safety records [over countries with different conditions], then comparing events per aircraft or flying hour is not necessarily a valid technique ... without taking into account many other significant factors”.
Yes quite. This brings in the second stage - modelling. When the common sense is implemented in mathematics, this field is called “analysis of variation”. One measures the overall variability of the data, then applies a candidate model, and recalculates the residual variation. If it is much less, then the model was a good one.
That is what I did. Using only the traffic at each aerodrome (pretty obvious), there was a clear reduction in variation, so that is an important factor.
Then I applied my second hunch – congestion at the boundary – and, after tuning, this reduced the variation further, so that factor is likely to be a real contributor also.
This does not rule other relevant factors - but they would belong in a more detailed layer.
Finally, this model is, I hold, the best estimate for the future, and should be taken as the “baseline MAC rate”. Being about half the Ambidji figure, it (along with the other errors) overturns their conclusions.
Leadsled #273 wrote:
“…one could easily get the impression that MACs were the biggest air safety problem in Australia.”
Yes, the Ambidji report reads as if midairs were the ONLY problem in Australia – because their brief was to investigate ATC under GAAP, which clearly affect mid-airs but less other accidents.
Their problem is that in they go on to draw conclusions about the total individual risk (to each pilot), and the total societal risk (to all pilots collectively) – after considering only one contribution.
They do not even mention single-aircraft accidents which are much more frequent. Nor do they mention ANY accidents outside GAAP procedures. No do they mention risk to third parties, ie those on the ground.
For example, according to Aviation Statistics, there were 1896 single-AC accidents in the last 10 years, but only 24 MACs. (Unfortunately their breakdown does not extend to fatalities.)
Leadsled wrote:
“Some time ago, a study was commissioned, using well known insurance co. investigators, and independent (of underwriters and the aviation sector) statisticians…”
Do you have a name or reference for that please?