Hey all -
Nothing to write up here, so..
Risk and return: I remember thinking for a long time that risk and return are just numbers, such as for an investment which returns 8% with standard deviation of 12% (i.e. the conventional measure for risk). I would go: “Ok.. so that means 68% of returns fall within -4% and 20%, while 95% fall between -16% and 32%…” But frankly this is super unintuitive: it doesn’t give much clarity at all on whether I can expect a -16% return or a 32% return, yet those are really different. Jeremiah Lowin has the best way of conceptualizing risk and return that I have ever heard: “return is a draw from the distribution that risk is trying to model” [44:34]. Now, we have a very clear picture of what we’re talking about: a distribution. It may be a normal bell curve, but it may also be lognormal or power law-shaped. All we have to do is visually look at our distribution to get a sense of the returns we could possibly draw. Fat tails imply a nontrivial chance of outlier outcomes, while long tails stretching far below zero imply an extremely small risk of absolute ruin (i.e. “black swan”). The great thing about visualizing risk this way is that is captures all the things we care about when we talk about risk: standard deviation, skewness, kurtosis, value at risk and so on. (surfaced by: Patrick O’Shaughnessy)
Statistical significance: The canonical example for statistical significance goes like this: say we have 200 sick people, then randomly give 100 medicine and the other 100 a placebo. After a week, 60 people from the treatment group get better while only 30 from the placebo group get better. While it seems like the medicine works, it’s possible that this particular experiment just happened to randomly put a lot more people in the treatment group who would have gotten better anyway, despite the medicine actually having no effect. Statistical significance measures how likely this difference (between control and treatment) is to have occurred by mere chance. For example, it would be extremely surprising - that is, unlikely - if all 100 randomly selected people for our 100-person treatment group happened to get better despite the medicine actually having no effect, while only 30 in the control got better. This is why the p-value is sometimes called the “surprise index” - how surprised should we be if there’s actually no effect? All that said, actually calculating your p-value is fraught with so many assumptions and mathy traps that it’s easy to lose the intuition: John Rauser explains p-values way better than I ever could in just 2 minutes. (found via: Ian)
Hyperglycemia and insulin-resistance: This is a 15-minute TED talk by McKinsey-alum and practicing physician Peter Attia. Peter makes two major points. (1) While it’s traditionally believed that obesity may cause diabetes (i.e. insulin resistance or shortage), Peter suggests that the causality may be the other way around: insulin-resistance may cause obesity. He hypothesizes that cells - flushed with too much glucose from a diet replete with breads, pastas & refined sugars - protect themselves from processing more glucose by rejecting the agent which processes it: insulin. As the body’s ability to process glucose deteriorates, more and more excess glucose gets stored as fat for a rainy day, hence obesity. If true, then obesity may be a symptom of insulin-resistance and diabetes, in which case we should focus on our diet, not necessarily weight loss. (2) Notice I say “if true!”: Peter emphasizes that this is simply a scientific hypothesis. He admonishes us to remain skeptical of ideologues who favor one hypothesis over all others; instead we must challenge our theories with empiricism in order to identify the truth. (referred by: Andras)
Corporate diversity: Some people think that diversity for the sake of diversity alone can problematic because it’s hard to identify the point at which you’re sacrificing performance for arbitrary groupings. For example, do we want to diversify our team with people from every town in the country, or every book club, or include hateful or bigoted people because we’re missing those qualities? On the other hand, there is a lot of evidence that diversity improves team performance. Personally, I think you must maintain a minimum bar of qualities: a team should be composed of people who are respectful (ie. promote psychological safety), open-minded, socially adaptable and intelligent. The Netflix culture deck promotes a similar philosophy: as long as people embody their nine core values (sound judgment, communication, impact, curiosity, innovation, courage, passion, honesty & selflessness), then feel free to diversify. I would go one step further: once everyone meets this minimum bar of civility and performance, I think you should diversify as much as possible with individuals who have completely orthogonal preferences, histories, creative inspirations & skill sets. (discussed with: Aaron)
Jerry G. and I were catching up in our classic discursive style - jumping from personal lives to data science to networking - until we finally moved to the topic of this newsletter. He starts with: “what I’m about to say isn’t a criticism, but more about what you want to do…” - which immediately gets me thinking, here comes the criticism. I focus intently on what he’s about to write.
He goes on to give me a couple of pointers, but pointers isn’t even the right word, because it was mostly questions for discussion. Questions like “do you think your newsletter could use more structure?” or “how can you develop your voice more?” and “is it worth tightening your scope?”
As he typed, a big grin appeared across my face. I couldn’t help but think, this is exactly the reason I write. Here I am focused on the details - writing and synthesizing content - but completely ignoring the high-level stuff. Such direct and to-the-point feedback is absolutely the fastest way to get better at anything.
Thanks for reading,
 I’ve been asked about statistical significance in a couple of interviews and calling the p-value a “surprise index” works every time.
 Thanks to Andras & Aaron for feedback as well so far. Still working on trying to make this shorter.