DA40 crash in Russia

Any DA40 related topics

Moderators: Rick, Lance Murray

User avatar
Kai
5 Diamonds Member
5 Diamonds Member
Posts: 1355
Joined: Tue Jun 15, 2010 6:14 pm
First Name: Kai
Aircraft Type: DA40F
Aircraft Registration: XXXX
Airports: VTCY
Has thanked: 127 times
Been thanked: 99 times

Re: DA40 crash in Russia

Post by Kai »

rwtucker wrote: (a) how many Lycoming DA40's are in the air
Good question. The serial numbers only might give us an idea how many have been produced but how many are still in the air after 16 (?) years of production?
rwtucker wrote: ... and (b) how many DA40NGs are in the air, and (c) how many VMC slow flight loss of control accidents have been attributed to each model (removing potential IMC causes from the equation).
Again, the serial numbers will give us an idea how many have been produced but no way we will be able to access all safety records of all countries the NGs have been sold to, especially as they have been sold to countries where safety records are harder to access than in the US.
DA40F - N405FP/HS-KAI (sold)
User avatar
rwtucker
5 Diamonds Member
5 Diamonds Member
Posts: 1283
Joined: Thu Feb 21, 2013 11:24 pm
First Name: Rob
Aircraft Type: DA40
Aircraft Registration: N831BA
Airports: KFFZ KEUL
Has thanked: 100 times
Been thanked: 110 times

Re: DA40 crash in Russia

Post by rwtucker »

There are ways to go about this; I didn't go into the details.

One way would go something like this.

You restrict your sample to data from the US, Canada, the EU and/or any other political entity that has readily available data. Diamond supplies year/month sold by model by country of registration. You create a "Model" variable for Lycoming and NG, discarding outliers, if any. You then index the resultant db on valid air worthiness certificates (not perfect but good) and create a "months of worthiness" variable for each record. You will lose some cases along the way but the bias is not likely to be systematic, more random. You then populate a variable for accident subtype: VMC loss of control (I am assuming that the EU database is similar to the FAA/NTSB; I know that Canada's is similar). The number of cases into which you are entering accident data will be so small that the new variable can be populated by hand, sourced with the appropriate federal data. You then run a series of simple significance tests using the "Model" field as a dependent variable. If Diamond will not supply the data (appropriately stripped of personal information), you work backwards from the available registration database(s).

These kinds of data sleuthing are done all the time. In doing them, you do not want to let the perfect become the enemy of the good. In cases like this, you might be able to get 90% of the way to 100% precision with a small fraction of the data and an equally small fraction of the overall effort.
Post Reply