Statistics is in many ways much more useful for most students than calculus. The problem is, to teach it well is extraordinarily difficult. It’s very easy to teach a horrible statistics class where you spit back the definitions of mean and median. But you become dangerous because you think you know something about data when in fact it’s kind of subtle.On replacing calculus with statistics | The Endeavour
From this repo. The network contains almost 9,000 bill cosponsorships between 559 MPs of the ongoing 14th legislature. The size of the nodes represents the weighted degree (the number of ties) of each MP with his cosponsors. The ultra-central Socialist is, unsurprisingly, Bruno Le Roux, the majority leader.
Don’t be a Nixon.
Comparing the way the Hagopian et al. survey has been presented, and the way the Roberts et al. 2004 Lancet survey was presented is also interesting. In both cases you have a central estimate of excess deaths with almost comical uncertainty surrounding it. For Roberts et al. this was an estimate of 98,000 with a confidence interval of 8,000 to 194,000. Then there is a public relations campaign that erases the uncertainty, leaving behind just the central estimate - 100,000 for Roberts et al. and 400,000 for Hagopian et al. Finally, the central estimate is promoted as a sort of minimum, with the “likely” number being even higher than their central estimate. Actually, Hagopian et al. went one step further, inflating up by another 100,000 before declaring a minimum of 500,000.MUSINGS ON IRAQ: Questioning The Lancet, PLOS, And Other Surveys On Iraqi Deaths, An Interview With Univ. of London Professor Michael Spagat
An alternative causal structure is that obesity harms everyone, even heart failure patients, and the lower mortality observed among obese patients with heart failure is an artifact of selecting the study sample from this subset of the population, in combination with unmeasured confounders of the relationship between heart failure and death.Commentary: Selection Bias as an Explanation for the Obesit… : Epidemiology (via hphwd, via CRS)
I do not think it is so helpful in most scientific settings to label null hypotheses as “true” or “false.” As we’ve discussed often enough on this blog, I’d much prefer to talk about Type S and Type M errors—that is, getting the sign of a comparison wrong, or overstating the magnitude of a comparison. Mistakes get published all the time—I’m with Val on that point—but I think it is helpful to to beyond the false-negative, false-positive thing.Statistical evidence for revised standards « Statistical Modeling, Causal Inference, and Social Science (which is why I tell students: you will get half the full grade for handling your nulls correctly, or all of it for handling your coefficients correctly; yet another way to say that useful results have meaningful units)
Academic equivalent of playing Street Fighter II in two-player mode. The Bayesians keep winning because the frequentist player base is larger and includes people who play no other button than “Start”.
(The data scientists play Minecraft instead.)
Legislative cosponsorships from this repo I just finished updating. The data are higher quality (removed a bit of noise, added MP details for session 12), the network plots are more flexible, and the centrality measures are now properly weighted.
The plotting function is ggnet, which should get a light update soon-ish to support additional layout parameters.