Lies, Damn Lies, and Bias

I enjoyed the episode. It certainly is amazing how easily biases get introduced into what you would think would be cut and dry (or controlled) situations. I am fan of Space and the US Space program, and in the middle of listening to the second season of a BBC podcast “13 Minutes to the Moon”. The Chernobyl example was interesting, and it reminded me of episodes 2 & 3 of that podcast. https://www.bbc.co.uk/programmes/w3ct0pc6 Those episodes, in particular, do a great job following Apollo 13 from the moment after the explosion showing how mission controllers were (in the urgency of the situation) not relying on the data in front of them. It took one controller who asked the questions that no one else was asking, and used those answers to say the data is correct and it was a real problem. The controllers initially didn’t understand that the data was in fact telling them that one of the titanium sphere liquid oxygen tanks burst.

With regards to data, the YouTube Channel 3Blue1Brown recently released the first two videos in a series he is doing on Probability of Probabilities breaking down the math on looking at quality data vs quantity. (example: What is better: 100% out of 10, 96% out of 50, or 93% out of 200) https://www.youtube.com/watch?v=8idr1WZ1A7Q&list=PLZHQObOWTQDOjmo3Y6ADm0ScWAlEXf-fp

There is definitely a lot that goes into data quality

Not a question for Josh as much as a rant I am hoping he will tone down and communicate better than I: I would appreciate if Josh would spend some time explaining to the world how machine learning and curve fitting are not models. A model predicts things that may not have been observed yet. Like the power required to roll at a certain speed on a bicycle. It’s based on principles of physics and engineering that have been verified and codified. The IHME model for the coronavirus is one such “model”. It does not include a mechanism to describe disease transmission or infection likelihood. It merely attempts to extrapolate off the data from prior disease spread in other cities, using “features” that are observed. Data can be used to help develop models. Curve fits are useful and can be used to estimate numbers, but they are not models. Without incorporating or explaining a phenomenon, graphs are not models. Hoping, Josh would like to make this point as much as I do these days.

John

Hey, I tried to post a comment before, but it disappeared from the site. Anyhow – I’m a randonneur, and the typical randonneuring style of bike has a big handlebar bag in front. That’s not my style, so I carry gear differently. But listening to this show reminded me that I’ve seen articles where Jan Heine of Bicycle Quarterly asserts that big front bags are actually more aero. Is that possible or is he just blowing smoke because he likes them? They’re big boxes!

Please clear up what’s likely a misunderstanding on my part. On this podcast, I’ve heard the “rule of 105” to refer to the ideal ratio of rim width to tire width regarding aerodynamics. That the rim would be wider than the tire surprised me, so I’ve re-listened and would swear it’s been described as rim being 105% of tire and not the other way around.

Cycling Tips just published an article on the new Zipp 303-S aero wheel. It features a 27mm rim “optimized around tires with a 28 mm measured width.” Doing the math, that equates to 103.7%. Please set me straight: does the rule of 105 refer to tire-to-rim ratio or rim-to-tire ratio?

“I have steadily endeavoured to keep my mind free so as to give up any hypothesis, however much beloved (and I cannot resist forming one on every subject), as soon as facts are shown to be opposed to it.” — Charles Darwin

I’ve known guys whose certainty in their own hypotheses increases not with how much they know but how little they think you know. I try to stay away from those guys.