Space Shuttle Challenger Disaster Caused by Bad Statistical Reasoning

Bad statistics can be deadly. On January 28th, 1986, 73 seconds after it launched, the space shuttle Challenger suffered a catastrophic failure and exploded, killing all 7 crew members aboard. This was one of the worst events to befall the space program and the loss of life was absolutely tragic. Unsurprisingly, a massive investigation was undertaken to understand what went wrong. This Rogers Commission concluded that the disaster was a result of unusually cold weather on the day of the launch that resulted in a set of O-rings failing to seal in the solid rocket boosters, which resulted in significant aerodynamic destabilization and ultimate explosion. But the report went further to conclude that one of the most telling failures wasn’t one of engineering, but rather the poor use and understanding of statistics.

Welcome to Data Demystified. I’m Jeff Galak and in this episode I’m going to show you how a failure in statistical reasoning by some of the smartest scientists and engineers in the world led to the tragic Challenger Disaster. In particular, we’re going to focus on how selective use of data can dramatically influence the conclusions we might draw.

Follow me at:

Equipment Used for Filming:

[shyama]

Nice touch in the end - the pic of challenger crew. Often in data analysis we tend to forget the real world and get caught up in jargon.

[scottm]

My friend here, although intelligent, has missed mark I'm afraid. Sure, because te engineers were so rushed that presenting their hurried data graphs probably were not the best. But this accident didnt happen because of bad presentation over safety. This happened because people were willing to roll the dice for political reasons and ignore the known issues that had been going on for years.

[anasxd]

Crystal clear explaination. Thank you so much

[ryangarbe]

Great video just showed it to my Intro to Statistics class for the concept of extrapolation.

[stevendegiorgio]

Thats all true, but Rochard Finmen proved that the O rings would fail in very cold temperatures.I believe if the Thiocol engineers did a physical experiment with the O rings in cold temperatures and the management at NASA saw it, they would have delayed the launch until warmer weather.

[zahraahmad]

rip to the legends. you reached great heights. 😇🙏🏾

[psychotic_madman]

so relating to this, what is your approach to censored data? the video makes it clear removing data leads to spurious conclusions, but what about datasets where limits of detection skew the data. reading around I see the substitution of LOD with 0, LOD/2, or LOD/sqrt(2). but have also read the effect of these substitutions can dramatically skew the data depending on the censure rate. One suggestion I did see was a mixed model treating data as binomial distribution for LOD, and then lognormal for those values above LOD (although not seen how to implement this)

[mikes]

From what i know, the engineers at thiokol were against launching

[notcharlotte]

To me, like a different angle of "base rate neglect", mentioned in one of your other video

[RatedArggg]

I don't know anything about statistics, but you'd think COMMON SENSE would have prevented them from launching.