Women and minorities represent less than 10% of pilots, yet were factors in four out of eight crashes (50%).

[candycorn]

Middle Aged White Dudes were responsible for 100% of deaths of NASA astronauts. Much higher than the 50% threshold in the OP, right? 100% higher!!!!!!!!!!!!!!!!!!!!!!!!

Therefore, I think it's time we give blacks, Hispanics, and women a chance to run NASA.

 

[scruffy]

Do you want to calculate the actual probability on this? We can do it together, it's easy.

First we'll need to answer two questions:

1. Are men and women equally capable and equally trained?

2. Is there any relationship between one crash and the next?

If we have the fraction of women pilots and the total number of pilots (we'll just ignore the transgenders lol), and the total number of flights and crashes, we can compare results from several different models, using several different statistical methods. If we know the date of each crash we can even predict the date of the next one.

To begin with I propose this model:

There is a parameter r, call it the rate of crashes. We would like to do two things in order: first, assume that r is the same for men and women, and estimate it based on the data. Then, assume that r is the rate for men only, and r' is the rate for women only, and estimate them both based on the data. Finally, look at the confidence intervals and see if the second model is any better than the first.

 

[NewsVine_Mariyam]

Your AI search fucked up. It was mentioned in the OP artice, but you missed it because you relied on AI.
Only sloppy and lazy journalists rely on AI.

You obviously do not understand how computer algorithms work. I explained to you exactly what criteria I used and you didn't raise any issues with it until after the fact. Atlas Air did not show up on my list because they don't carry passengers, they transport cargo.

So the thing is, no matter which way you attempt to present this information, it still DOES NOT equate to those bogus numbers Huff came up with and that you swallowed hook, line & sinker presumably because it supports your biases.

For passenger carriers there are no crashes caused by either a female or a minority. Adding in the Atlas Air incident STILL doesn't get you to what you claim - Aska, the Black first officer was not the pilot in command it's the captain who ultimately is responsible for the safety of the aircraft, the flight crew & passenger as well as people on the ground. While a 1st officer may be reluctant to challenge the captain, there should not be any reservations on the part of the captain in challenging his or her first officer or more importantly taking control from them should they veer from proper procedure & training.

Again, this single incident still does not produce nor support the claim that 50% of fatal commercial airline accidents are caused by only 10% of the female/minority pilot workforce, particularly since the math Huff used to formulate his percentage is also bogus.

I mentioned when I first responded to your OP, that I didn't know why he used the year 2000, but now I do. That produces a really small denominator of 5 for passenger carriers or 6 for passenger & cargo. That means that just 1 incident would result in a 20% crash rate for fatal passenger carrier crashes or 16.7% rate for combined passenger & cargo which gives a false impression of overrepresentation. But that's not the proper way to calculate the numbers. To properly calculate raw numbers without attribute you'd have to start with the first year that Black pilots began flying for the commercials airlines which was 1963, Marlon Green who was hired by Eastern Airlines.

Either way the numbers are calculated, whether honestly or disingenuously, why planes crash is generally a multi-faceted chain of events and the NTSB was one of the very first organizations whose investigative reports I began reading, cover-to-cover, their "blue books". So I at least have some background to support my claims that Huff's "research" is suspect.

 

[NewsVine_Mariyam]

Do you want to calculate the actual probability on this? We can do it together, it's easy.

First we'll need to answer two questions:

1. Are men and women equally capable and equally trained?

2. Is there any relationship between one crash and the next?

If we have the fraction of women pilots and the total number of pilots (we'll just ignore the transgenders lol), and the total number of flights and crashes, we can compare results from several different models, using several different statistical methods. If we know the date of each crash we can even predict the date of the next one.

To begin with I propose this model:

There is a parameter r, call it the rate of crashes. We would like to do two things in order: first, assume that r is the same for men and women, and estimate it based on the data. Then, assume that r is the rate for men only, and r' is the rate for women only, and estimate them both based on the data. Finally, look at the confidence intervals and see if the second model is any better than the first.

Your model doesn't have enough features to allow it to generalize correctly, therefore it will perform poorly.

And that's aside from the fact that the NTSB doesn't consider race or gender in its investigations:

In its accident reports, the National Transportation Safety Board (NTSB) determines probable cause based on evidence related to aircraft systems, crew actions, training, procedures, weather, air traffic control, and organizational factors. The agency does not list race or gender as causal variables in its investigative methodology or probable cause determinations.

 

[scruffy]

Your model doesn't have enough features to allow it to generalize correctly, therefore it will perform poorly.

Well, my answers to the questions are Yes and No, in that order.

Given that information, we have two choices right off the bat. First, we can treat each flight as a Bernoulli trial, like a coin flip. It either crashes or it doesn't. The probability of it crashing is p, therefore the probability of not crashing is 1-p (there are only two choices). We expect p to be small (hopefully), but we don't really know for sure, so what we do is find the value of p that best explains the data. The method is called "maximum likelihood estimation".

A second way we can look at the data, is as distance between crashes. In this model crashes are random, the null hypothesis is they happen because of mechanical failure and have nothing to do with gender or race. This version is a little more complicated because there are more options. The basic statistic is called a Poisson distribution, it measures the time between events. It has a single parameter lambda which is the rate (or time, they're reciprocal). You can relate it to the Bernoulli model using a beta distribution, by specifying number of successes and number of failures. So like, the bus either arrives or it doesn't, within a given time window.

And that's aside from the fact that the NTSB doesn't consider race or gender in its investigations:

Well, so you have to do the math. First you test a single rate without considering race or gender, then you test the other model, with two different rates, one for men and one for women. You have to look at the mean square of the variances to figure out a confidence level. In the model that doesn't fit (or if you don't have enough data), the variance will be much larger than the mean, and the confidence levels will be very low.

 

[NewsVine_Mariyam]

Well, my answers to the questions are Yes and No, in that order.

Given that information, we have two choices right off the bat. First, we can treat each flight as a Bernoulli trial, like a coin flip. It either crashes or it doesn't. The probability of it crashing is p, therefore the probability of not crashing is 1-p (there are only two choices). We expect p to be small (hopefully), but we don't really know for sure, so what we do is find the value of p that best explains the data. The method is called "maximum likelihood estimation".

A second way we can look at the data, is as distance between crashes. In this model crashes are random, the null hypothesis is they happen because of mechanical failure and have nothing to do with gender or race. This version is a little more complicated because there are more options. The basic statistic is called a Poisson distribution, it measures the time between events. It has a single parameter lambda which is the rate (or time, they're reciprocal). You can relate it to the Bernoulli model using a beta distribution, by specifying number of successes and number of failures. So like, the bus either arrives or it doesn't, within a given time window.



Well, so you have to do the math. First you test a single rate without considering race or gender, then you test the other model, with two different rates, one for men and one for women. You have to look at the mean square of the variances to figure out a confidence level. In the model that doesn't fit (or if you don't have enough data), the variance will be much larger than the mean, and the confidence levels will be very low.

Okay so what will the probability tell us? In plain language, no pun intended

 

[scruffy]

Okay so what will the probability tell us? In plain language, no pun intended

Well, the outcome will be the confidence level of a hypothesis.

In this case the hypothesis is that race and/or gender make a difference. The null hypothesis is that they don't.

So we want to see if the data is strong enough to reject the null hypothesis. Generally the statisticians use 95% as a confidence boundary.

 

[NewsVine_Mariyam]

Well, the outcome will be the confidence level of a hypothesis.

In this case the hypothesis is that race and/or gender make a difference. The null hypothesis is that they don't.

So we want to see if the data is strong enough to reject the null hypothesis. Generally the statisticians use 95% as a confidence boundary.

Have you ever actually read an NTSB accident report? I ask because they’re extraordinarily detailed and multidisciplinary. Investigators examine aircraft systems, maintenance history, crew training records, cockpit voice data, flight data recorder data, weather, ATC communications, company procedures, fatigue, CRM dynamics, regulatory compliance, and more.

Reducing a crash to a binary variable like race or gender ignores how aviation safety analysis actually works.

Commercial pilots operate in one of the most standardized, regulated, and audited professions in the country. To reach the left seat requires thousands of flight hours, recurrent evaluations, simulator checks, line checks, and adherence to federal standards. Suggesting that crash causation can be meaningfully modeled around demographic traits — especially from a dataset of only a handful of events — misunderstands both aviation and statistics.

The NTSB does not consider race or gender in its causal determinations because accident causation is systemic and operational, not demographic.

 

[GLASNOST]

Women and minorities represent less than 10% of pilots, yet were factors in four out of eight crashes (50%).​

That's bad odds!

 

[Dorian Mode]

View attachment 1221116

That's bad odds!

What percentage of airline pilots are asian women? Isn't that like the worst possible nightmare behind the wheel of anything that moves? If I am about to board a plane and see an asian woman as the pilot, I turn around and catch the next flight.
.
.
.
.
Just having tongue in cheek fun!

 

[GLASNOST]

What percentage of airline pilots are asian women? Isn't that like the worst possible nightmare behind the wheel of anything that moves? If I am about to board a plane and see an asian pilot, I turn around and catch the next flight.

Oh, my goodness!

Just having tongue in cheek fun!

Ah yes, I see now.

 

[Dorian Mode]

Oh, my goodness!

Ah yes, I see now.

Ah, I had to fix my post to be more clear! Anyway I always pick on my Italian wife about her driving. I nicknamed her "Crash." She always ask "Guess what?" I immediately respond "You crashed the car again?" and then she kicks my ass.

 

[scruffy]

The NTSB does not consider race or gender in its causal determinations because accident causation is systemic and operational, not demographic.

No, that's an opinion on your part. Willfully ignoring data is DEI-think, it puts peoples' lives at risk. Pilot background may very well be a factor in crashes. We want to know how these people did in school too. And who certified them. Especially after that guy got busted in DC, handing out test answers to applicants of color.

 

[GLASNOST]

Ah, I had to fix my post to be more clear! Anyway I always pick on my Italian wife about her driving. I nicknamed her "Crash." She always ask "Guess what?" I immediately respond "You crashed the car again?" and then she kicks my ass.

 

[toobfreak]

The NTSB does not consider race or gender in its causal determinations because accident causation is systemic and operational, not demographic.

1. The NTSB does not need to consider demographics in an investigation. For one thing, they may not be allowed to.
2. It is all a moot point. All one needs to do is look at the occurrence of pilot-error related crashes, then see if the occurrence of minority-based or female-based accident involvement exceeds their percentage of participation.

If a race or gender only occupies, say, 25% of all pilot positions yet are involved in 50% of all crashes, you've shown an unhealthy frequency of occurrence of pilot errors for that race/gender above the mean norm.

If neither race nor gender were not any factor in pilot-error related crashes, then if only 10% of all pilots were minority or female, then only 10% of them would be involved in pilot-error related crashes.

While the NTSB might not track such a thing openly or publicly, it is very likely that at some level, someone does.

 

[WorldWatcher]

1. The NTSB does not need to consider demographics in an investigation. For one thing, they may not be allowed to.
2. It is all a moot point. All one needs to do is look at the occurrence of pilot-error related crashes, then see if the occurrence of minority-based or female-based accident involvement exceeds their percentage of participation.

If a race or gender only occupies, say, 25% of all pilot positions yet are involved in 50% of all crashes, you've shown an unhealthy frequency of occurrence of pilot errors for that race/gender above the mean norm.

If neither race nor gender were not any factor in pilot-error related crashes, then if only 10% of all pilots were minority or female, then only 10% of them would be involved in pilot-error related crashes.

While the NTSB might not track such a thing openly or publicly, it is very likely that at some level, someone does.

That's not true. Techically.

You eliminate experience from the equation. Since the commercial pilot industry was dominated since inception by white males. You would need to factor in both:

• Experience as in total flight time, and
• Experience as in flight time in model

So not only would the need exist to correleate race and/or gender to previous accident data, you would need to do it as a function of (at least) two type of experience. And then those would need to be differentiated by Polit in Command and Co-Pilot.

WW

 

[scruffy]

You guys are making it too complicated.

We want a simple yes/no answer. Was gender a factor? Was race a factor? Yes or no?

If we say yes, we have to be able to do so with 95% confidence.

I don't have the data, if someone wants to point me to it I'll do the math.

 

[scruffy]

If neither race nor gender were not any factor in pilot-error related crashes, then if only 10% of all pilots were minority or female, then only 10% of them would be involved in pilot-error related crashes.

This is exactly like a drug trial. We have a drug called "race" and it's going to make you crash. But it only works in a small fraction of the population. You get false positives and false negatives, and you have to have enough data points to overcome the uncertainty.

 

[Admiral Rockwell Tory]

What percentage of airline pilots are asian women? Isn't that like the worst possible nightmare behind the wheel of anything that moves? If I am about to board a plane and see an asian woman as the pilot, I turn around and catch the next flight.
.
.
.
.
Just having tongue in cheek fun!

How do you blind an Asian driver? Put them behind a steering wheel!

 

[Carl in Michigan]

Carl in Michigan