Skip to main content

Survivor-ship Bias: Why everyone's advice (including mine) is useless.

There is a pseudo-mythic story about Abraham Wald, a Hungarian mathematician, and the US Air Force (USAF) that I'd like to relate to you.  There's some question as to how much of this story is fact, and how much is embellishment, but it's close enough to true to be useful for my purposes.  

By the time the US got involved in WWII, the Germans were pretty well entrenched in the lands they occupied.  Part of this entrenchment involved land based anti-aircraft batteries.  These anti-air defenses wreaked all merry havoc on US bombers.  In attempt to address this, the USAF turned to the Statistical Research Group at Columbia University, where Wald was a member.  They wanted an analysis of where their planes were being hit, based on where the holes were in the planes they got back.  The statisticians went to work, and found a locations that had statistically many more holes than other locations.  (I would be remiss if I didn't tell you at this point that I have no idea what analysis they did, or what their criteria were.  It's not relevant to the story, and I don't care to send time looking it up).  The USAF concluded that if those area were being hit more, then those where the areas they needed to more heavily armor and reinforce.  This was absolutely the wrong conclusion.  If any of you have figured out why, now is the time to pat yourself on the back.  The reason this is wrong is because of a concept called "survivorship bias."  They only had data from the planes that made it back, so those density maps?  They showed the places a plane could be shot all to hell and keep flying just the same.  They're the last places you need to reinforce.  Now you may realize that this data still has some predictive ability.  You just have to look at the places where these planes were NOT hit, and armor those (as those represent the places were the planes that went down probably were hit), but you can only do that when you recognize the bias and the full data set.  

What does that have to do with us?  Quite simple.  It means that any advise I, or anyone else has to give you is probably worthless because it is bias towards what worked for that person, in their specific situation.  For example, there is a famous commencement speech where a US Navy Admiral advises a class to "make your bed every morning."  Now, you have no reason to know this, but I am NOT a morning person.  It's a good day if I can stumble from my bed to the bathroom sink without walking into a wall or a door-frame.  I never make my bed in the morning.  Most days, the best I do is pick up any blanket or pillow I might have kicked in the floor overnight.  Most days, not all.  That particular piece of advise does not work for me.  It worked for him because of his background his experiences made that something meaningful, because in the military making up your bunk is a task that MUST be done, no option.  Because it had to be done no matter what, it became an accomplishment for the day.  For me it's not.  I don't have to make my bed, and I derive no sense of accomplishment from doing so.  Now, I still encourage everyone to watch that speech, because there is wisdom in it.  Adml. McRaven goes on to explain that for him, making his bed in the morning is that first little accomplishment of the day.  That one thing leads to more, and whatever else happens, you've got at least that one thing done.  It's an idea I can get behind.  It's just that for me, that first accomplishment is not making my bed.  

What I am saying is that every successful person will say "I did this and it was the key to my success," sometimes adding "if you do this, you'll be successful too."  They can be telling the truth about their success and still be dead wrong about yours, because what was key for them was a result of the selection bias of their situation.   

If that's the case, then why bother with advise?  Why bother listening?  Why do I bother writing all of this for you to read?  Because there is wisdom in experience, and sharing that with each other betters us all.  But all knowledge and all wisdom both have to be worked for.  You have to find the truth or core idea of any piece of advise and decide whether or not that works for you and how.  It's some of like with anything in science, you have to be open to every idea, but also you have engage with it critically to make sure it works.  For me, the idea that getting that first thing accomplished in a day leads to more is something that really helps me when I have a long "to-do" list, but that first things sure as hell ain't making my bloody bed. 



Figure out what works for you, do it by looking at the data,
Faxe MacAran
Twitter: @TheMacAran

Comments

Popular posts from this blog

Please Consider a Pot Roast

To quote Stephan Pastis's Pearls Before Swine : How many roasts must a man wok, Downs, before he is called a nan?  If you're unaware it's a pun playing off the question "How many roads must a man walk down before he is called a man?" which Bob Dylan used in his song "Blowing in the Wind" as a philosophical statement.  Yes dear reader, I I did just explain the joke, you're just going to have to deal with it.   So, what's all this rambling mean about grad school?  Well, not much, other than that the pun fits well into the analogy I want to make.  Consider a pot roast.  They make fantastic meals, but they also take a fair amount of time to cook.  Generally, you follow the rule of about an hours cook time per pound.  So a 4lb roast takes 4 hours in the crock pot.  There is no speeding it up.  It takes as long as it takes.  And that idea does have a lot to do with grad school.   Anyone who's ever taken a che...

Classes? We don't need no STINKIN classes!

Qualifications? Misrepresenting statistics.  Misrepresenting statistics? That's not much of a crime.  During a public health crisis.  Kinky... sign here. 1 Ok, that joke might be hitting just a little to close to home, but I was watching Blazing Saddles this weekend and wanted to work it in somehow.  Anyway, 5 months in and I'm finally getting back in the lab.  Brushing off rusty skills, being incredibly paranoid about what stuff I might have forgotten to do, and overall feeling like that gif of a dog in a lab coat hoping no one notices.  But that's ok right?  I've got time to get back into practice without a bunch of distraction and get back into the swing of things right?  Wait, what do you mean it's time to register for my fall courses?  Oh right, those things.   At this point in my academic career, I have a mixed relationship with coursework.  On the one hand, courses are a good way to learn new material, and also new ...

Theory and Practice: When nothing works and no one knows why.

There's a joke I've heard time and again.  "Theory is when you know everything but nothing works.  Practice is when everything works but no one knows why.  In our lab, theory and practice are combined: nothing works and no one knows why." I have no idea where it originated, but I've seen it printed out and posted in at least 6 different labs that I've either worked in or visited.  It's a nice play on words, but it's also eminently relatable to everyone who's worked a lab. At the moment, I'm rotating in a cell culture lab.  We grow undifferentiated cells in a culture dish until we have a large enough colony, and then differentiate them into vascular tissues.  We can try a range of drugs, inhibitors, growth factors and other treatments to the cells to see how it changes the tissue growth.  There are four of us doing cell culture right now: myself and one of my classmates, a post-bacc student, 1 and a more senior grad student who has been teach...