I don’t pretend to understand this. Maybe after three or four readings I’ll get the gist, but I think this lays bare the truth behind “There are statistics, damned statistics, and lies”.
Phew. The Stata software is being singled out here for algorithmic compensations that could create a bunch (more) of statistically normal results. I know that this app is used extensively by quants in the field of economics. I’m more familiar with Matlab.
Matlab is kind of like the training wheels on a bike. Produces lovely colourful 3D figures that look great on a PowerPoint. Much cheaper to license… considered PoundShop software by the economics quants.
They are the overwhelming majority these days because they produce results, but even more so econometric models, that are measurable and capable of being monetised. Count those lovely grants. Meanwhile a few stubborn pure economics theorists cling on in the field but I wonder for how much longer. (Psychology is the same: they barely even touch upon Freud et al these days: too old, let’s spin up some cool spreadsheets instead.)
The whole field of Economics is based on “…let’s assume [whatever]” and creating straw men from the results. Long statistical runs on a juicy GPU, days long, are like flashing your Strava mileage figures as proof of how rock-hard you are. (Econometricians are broadly speaking all males.)
If the app “smooths out” certain jagged bits of data I expect that other packages do the same.
The example used by the author relates to voter preferences and seeking to pin down variability of political sentiment in a model voter… but this is where the discussion lost me.
Lies damned lies and psephology
@Kieran_Telo I got lost a bit before that, thanks for the exposition.
Yeah @PatB, randomized controlled trials or RCT have emerged as a leading ideological tool, weapon even. As they need to be large they are the dominion of industry, pharma and governments. Very handy then, for dissing the studies of independent groups that show inconvenient results. The brazen propagandizing heightened during covid (especially with respect to hydroxychloroquine) when, in full view of the world’s scientists, it emerged that any number of careful, perfectly valid, independently conducted observational studies were (with the help of the ‘science’ media) trumped by even a single randomized trial - even if that was a terrible study, studied what was essentially a different thing or if the wrong variable was randomized or the study was conducted in a clearly dishonest fashion. The words ‘randomized controlled trial’ became a mantra to stifle opposition.
This is a long but telling article about Hcq; I’ve picked out some RCT bits
ED
Randomized controlled trials
Philipe Rafaeli
"In a 2011 a scientific study by Dong Heun Lee showed that only 14% of the Infectious Diseases Society of America’s treatment guidelines were based on the “highest level” of evidence, the requirement they want from hydroxichloroquine.
When we talk about cardiology medicines, 89% of the recommendations do not have this evidence.
Have you ever seen a doctor who prescribes insulin for diabetics called an “anti-science” or a quack? They should be called that, because there has never been a “gold standard” study on insulin.
Have you ever been vaccinated against tetanus? It also does not have the “highest level” of evidence. Should we say that this vaccine was made without scientific basis? The theorists of the anti-vaccine conspiracy would go to the delirium.
It was never done for insulin or tetanus vaccine because it is not necessary. In a 2014 scientific paper, Andrew Anglemeyer, an epidemiologist at the University of Otago, New Zealand, explained that there are no significant differences in results between observational studies and the “gold standard”. In another study, published in 2000, Kjell Benson came to the same conclusions.
In other words, historically, the results shown by the observational studies match the “gold standard”.
Moreover, before the pandemic, science used to say something else. Angus Deaton, a scientist at Princeton University, USA, in 2018, explained that the gold standard is not as “gold” as they say. “RCT results can serve science but are weak ground for inferring ‘what works’”, he explained.
In a 2017 article, the scientist Thomas Frieden, appointed director of the U.S. CDC by Barack Obama, argued in his scientific article, published in the New England Journal of Medicine, that public decisions on treatment should not be based on RCTs.
Frieden commented on the “increasing costs” of “gold standard” studies, in addition to the delay in carrying out such research. “These limitations also affect the use of RCTs for urgent health issues, such as infectious disease outbreaks,” he wrote.
In recent years, just before COVID-19, medical science seemed quite reasonable, didn’t it?"
Gold Standard is exactly the phrase used across the social sciences, yes. Has a hypnotic effect and the credulity threshold, if it was ever very high, plummets.