lukestein’s Twitter Archive—№ 13,200

The best teaching moments often come when a student asks a question and with the resulting discussion you can just see the class’s lightbulbs turn on We had just generated this graph and the question was: “Shouldn’t 95% of the dots be in the shaded areas? They’re not.” 1/
Permalink On twitter.com ♻️ 5 Retweets ❤️ 23 Favorites 2021 Nov 30 Mood +3 🙂

…in reply to @lukestein
I got to use one of my favorite examples as a pedagogical light switch: Suppose I flip a possibly biased* coin 10 times and we see 6 heads. What is your prediction for the next flip? How confident are you in your prediction? 2/
Permalink On twitter.com ♻️ 1 Retweets 2021 Nov 30 Mood +2 🙂

…in reply to @lukestein
Now suppose I flip the coin 10,000 times and we see approximately 6,000 heads. What about now? 3/
On twitter.com ♻️ 1 Retweets ❤️ 1 Favorite 2021 Nov 30 Mood 0

…in reply to @lukestein
One fun thing is there’s always a (wise? wise ass?) student who says we can’t predict after the 6/10, since the coin is probably fair. A Bayesian prior! Of course after 10K we’re much more sure p≈0.6 but not much more confident in our prediction that the next toss will be H. 4/
Permalink On twitter.com ♻️ 1 Retweets ❤️ 3 Favorites 2021 Nov 30 Mood +4 🙂

…in reply to @lukestein
I ask MS Finance students to interpret a CAPM beta in a sentence, and it seems like in undergrad courses they somehow *learned* (incorrectly) that beta < 1 stocks like $TWTR are “less volatile than the market.” 5/
Permalink On twitter.com ♻️ 1 Retweets ❤️ 2 Favorites 2021 Nov 30 Mood +2 🙂

…in reply to @lukestein
Uh, @Investopedia: “Companies that are less volatile than the market have a beta of less than 1 but more than 0… A beta of 1 means a stock mirrors the volatility of whatever index is used to represent the overall market.”investopedia.com/investing/beta-gauging-price-fluctuations/ 6/
Permalink On twitter.com ♻️ 1 Retweets ❤️ 2 Favorites 2021 Nov 30 Mood 0

…in reply to @lukestein
I think we show them too many regression lines and not enough scatters/residuals. Also, * from above: Acording to @StatModeling there’s no such thing as a biased coin🪙🤷 7/ @lukestein/1053831095542669312
Permalink On twitter.com ♻️ 1 Retweets ❤️ 1 Favorite 2021 Nov 30 Mood -3 🙁

…in reply to @lukestein
@StatModeling P.S. We also, we show them too much pre-cleaned data 8/_N @lukestein/1462818939919126535
Permalink On twitter.com ♻️ 1 Retweets ❤️ 4 Favorites 2021 Nov 30 Mood 0