Survivor Bias is when you only consider data that meets a condition (such as surviving) and not correcting for that condition. By forcing data to meet a condition you are limiting your data. When doing research you have to limit your conclusion when you limit your data.
e.g.
Theory: homes built 200 years ago were more beautiful. Reality: Only the prettiest homes built 200 years ago survived because the rest were knocked down. Error: limited data to only surviving homes built 200 years ago and ignored the non-surviving ones. Corrected: The 200 year old homes I see today are beautiful”
Theory: We never had car seats when we were young and we grew up fine so car seats are unnecessary. Reality: not everyone did grow up fine. Error: not including major injuries or deaths due to lack of car seats. Corrected: “Thankfully I never got in a bad accident so I never had to test a car seat’s safety”.
Theory: They don’t make quality cars like they used to *pats 50 year old car*. Reality: Not many are left on the road. Error: Just because one (or some) survived doesn’t mean all cars built at that time could or would have survived or that they were built better. Corrected: “This car was built and maintained well or else it wouldn’t be here today”
This would be the simplest example I can think of
“Playing chicken on the freeway isn’t dangerous, we used to do it all the time and it did us no harm!”
… the kids who got hit by trucks and died, unsurprisingly, aren’t able to chime in with an “actually no, it’s quite dangerous, I died”
That’s it, that’s survivorship bias in a nutshell – you only see the success stories. Old buildings are beautiful and sturdy because the ugly and flimsy ones got knocked down and replaced. Old appliances are more reliable because the millions of unreliable ones broke decades ago and have been forgotten about
“I don’t understand why kids these days need to be in these fancy car seats every minute they’re in the car. We didn’t have any of that stuff and we were fine.”
Yes, those of you who survived to adulthood and can now rant against improved safety measures are fine. However, all those of your generation who died in car accidents aren’t around to tell their story, so the information presented is inherently biased.
The best response to such logic is “yes, you survived. But *fewer* of you survived”
Survivorship bias means you only look at successful participants when evaluating an entire group.
Example 1: Justin Bieber started a Youtube channel, posted videos of himself singing, got discovered, and became a millionaire pop star. By only examining the results of Justin Bieber, the “survivor,” it would appear that making it big in the music industry just requires starting a Youtube channel.
When you remove the bias and examine the hundreds of thousands (or millions) of aspiring musicians who created a Youtube channel, you see that the rate of success is extremely low.
Survivorship bias ignores most (or all) of the failures in a group and only focuses on the information about the individuals that succeeded, leading to incorrect conclusions.
I think it s easiest described with the War War 2 story in wich they tended to look at planes that did come back from the mission and see where the holes are.
At the beginning they started reinforcing those areas thinking that most aircraft gets bullets exactly in that region. That s the definition of survival bias.
Along the way someone smarter figured out though that because all the planes that did come back had holes in the same region and survived, the ones that did not must have them on the other parts of the plane. The surviving one clearly could cope with the holes where they were.
So they started reinforcing the ares that weren t full of holes with great success and increased survivability
Latest Answers