Tag: statistics

[Substack] Medical Mistakes with Probability, 2

On By In Academic, General, Statements

Abstract: risk factors for LDL/ApoB underestimate the risk factor for Lp(a) positive subjects and overestimate for Lp(a) negative ones, a case of base-rate fallacy.

I am just using basic probabilistic logic here.

Risks factors for ASCVD from LDL (or ApoB) levels are computed for a general population which includes people with low and high Lp(a) levels. Now if having a high Lp(a) increases the cardiac risk over the baseline (up to 2-3 times!) and the proportion of subjects with high Lp(a) is between 15 and 28% of the population, then, necessarily, those with low Lp(a) will have, for a given level of LDL, a considerably lower risk and many might be treated unnecessarily.

The risk factor for nonLp(a) can be ~ 30% lower! Statins don’t come for free. There are hidden and less hidden side effects.

Continue reading on Substack – open.substack.com/pub/nntaleb/p/medical-mistakes-with-probability-284

[Substack] Medical Mistakes with Probability, 1

On By In Academic, General, Statements

Max Heart Rate

When you do a stress test, say the Bruce Protocol, the administering doctor relies on something called the “age predicted maximum heart rate”, usually 220 minus your age, or some formula slightly more complicated but equally unrigorous. Once you reach that point, they stop, depriving you of potent information — at low risk since they are monitoring via live ECG your cardiac strain. In fact, such an estimation based on age, no matter how complicated its computation, appears to explain only 20% of the variation between individuals. I believe that explained variations are even smaller for, clearly, in the graph above, to the right, samples above 55 are sparce and the expected maxima would be considerably higher.

I noticed this myself as I am easily able to reach the 170s without feeling strain, guessing the effective max would be in the 180s (next test, but would require some live ECG for caution).

Continue reading on Substack – open.substack.com/pub/nntaleb/p/medical-mistakes-with-probability

[YouTube] MINI LECTURE 15 – Conditional vs. unconditional correlation: twin studies overestimate heredity

On By In MOOCs, Videos

The genetics of twin studies have a bias showing more heredity than in reality, owing to a statistical artifact. The twin studies for heredity are based on comparing the correlation between 2 identical twins minus that between 2 fraternal ones (assumed to be sharing half their genes). The use of fraternal twins as control is assumed to extract the “environmental” factors. Problem: Correlation is conditional and psychologists think it is unconditional. We show how the math is entirely different. The core error is that genes and environment are not separable and additive.