felis-parenthesis
No bio...
User ID: 660
felis-parenthesis
16d ago
That is a bad comment. You are replying to a comment that claims there have been twenty-four years of disasters and laments that people are not learning. Have there been twenty-four years of disaster? Does past performance predict future performance? Is this time different? There is plenty to engage with. But your comment is negative, low effort, and unresponsive.
felis-parenthesis
17d ago
I had a very negative reaction to this article. I think it reads much better with the following set up to contextualize it.
Start by contemplating the power of bench top science and theory. The partnership of bench top science and theory has some spectacular successes to its credit. You can experiment with Newtonian mechanics in your laboratory, verifying the basic laws. Then you get a top mathematician (Euler) to work out what those laws imply for gyroscopes. Later engineers build a gyrocompass for a submarine. Will this really work? Underwater? A thousand miles from the laboratory? Yes!
Or think of James Clerk-Maxwell, taking the bench top science of Ampere and Faraday and coming up with Maxwell's equations. The equations predict electromagnetic radiation. Hertz does the experiments in his laboratory and finds them; a great triumph for theory. Later Marconi takes this out into the real world. Theory shows that electromagnetic radiation goes in straight lines; Marconi's attempts at radio communication beyond the curve of the earth are not going to work. And sure enough they fail. Wut! Marconi actually succeeds! But rather than concede that there are problems getting out of the laboratory and into the real world, we discover the ionosphere and chalk it up as another spectacular success for the partnership of bench top science and theory.
Move on to contemplating the contrasting situation in medicine. The human body is too complicated for the human mind to comprehend. Basing medicine on theory works badly for the patient, no matter how much money it brings in for the doctor. This has lead to evidence based medicine. Never trust the combination of lab bench chemistry and theory. Always do a randomised controlled trial to check that medicines really work. Theory said that vitamin E was an anti-oxidant and would reduce oxidative stress and prevent cancer. Epidemiology confirmed this. Randomised controlled trials refuted it. Examples are so numerous that you can fill a book.
Returning to climate science, we must ask whether it is like gyroscopes or like Vitamin-E/Vioxx/vertebroplasty. Doing bench top science with an infrared spectroscope and a sample of Carbon Dioxide yields uncontroversial results. But does it have implications for the weather?
I've picked up the impression that every-one agrees on the importance of feedback loops. If you believed that climate science was a like a gyroscope, you would compute warming on the basis of the infrared characteristics of Carbon Dioxide and conclude that we are in for some warming, but not enough to constitute a crisis. No-one believes that climate science is like a gyroscope. Some think that warmer air means more water vapor which means more warming and more clouds and more clouds mean mumble. Subtle feedback in the atmosphere is putting us on a course for disaster. Others disagree.
The article emphasizes one particular disagreement. Scientists attempt empirical confirmation of the theory, but they mess it up. All the empirical work is heavily contaminated by theory. It cannot refute the theory because it assumes the theory.
felis-parenthesis
18d ago
I clicked your link and saw "Deleted by user" when logged in (as me, not you). So yes, there is problem. But I'm merely a peasant, tilling my freehold outside of the Bailey, so I cannot help :-(
You will sometimes see a medical study described as a "blinded randomized controlled trial" and other medical studies as an "open label randomized controlled trial". Whether a study is blinded or open-label is a separate issue from having a control group and separate again from having random assignment to treatment. The Wikipedia page on GRADE doesn't mention blinding. Checking the website of the National Institute for Clinical Excellence (NICE) the reference to randomized controlled trials does not mention blinding.
A study on breast augmentation can qualify as a randomized controlled trial, generating top quality evidence, with the following design: recruit 200 subjects. Randomly assign 100 to have the operation. The others are the control group. Researchers must keep in touch with the all 200 to find out how things worked out for them.
Keeping in touch with all 200 might be tricky. Some of those in the trial group might be disappointed with the results of the surgery and feeling disillusioned with medical intervention may reject further contact with the trial. Some of those in the control group may interpret being rejected for surgery as being rejected from the trial and disappear. Such people are lost to follow up. How to analyse a trial with large loses to follow up is controversial. Do we blandly say "we don't know"? Should we interpret losses from the trial group as bad outcomes? One might vary the design. 100 breast augmentations. 50 get psychotherapy that aims to persuade them that they don't need breast augmentation. 50 get regular contact to keep in touch, but bland contact, merely reminding then that they also serve who only stand and wait.
Scott's deep dive into Alcoholics Anonymous is my goto article for the practical importance of having a control group. Or should that be the disappointing effect of control groups?
If a trial does not have randomization, it is vulnerable to Simpson's Paradox. One may find that a medical treatment is beneficial, but partitioning the data into two exhaustive and mutually exclusive subgroups, find that the treatment is harmful to one of the subgroups and also harmful to the other subgroup. Wut? The analysis may collapse into baffling incoherence. Actually it is worse than that. The laws of arithmetic are chaotic evil, and permit that a conclusion that has been reversed by one partition may yet be swapped back by a finer one (if the Chrome browser objects to a faulty certificate, using incognito mode will work.)
The two issues, of needing a control group, and of needing randomization, are widely understood; I would not expect Dr Hilary Cass to restate the arguments in her report.
Edited to fix link to Simpsons Paradox, spotted error way too late :-(
In 2009 I suggested Saving forums from themselves with shared hierarchical white lists Linking to archived page: project name Outer Circle. It was discussed on https://news.ycombinator.com/item?id=920110 My health deteriorated. Nobody else tried implementing the ideas.
That allows for planning ahead and maybe writing your own version of Outer Circle. The core idea is
Shared hierarchical white-lists are a mechanism for allowing "multiple forums" to peacefully coexist with the same "comment base". You don't see shitty comments because you don't white-list them. The shitty commentators don't try to ban you because they never white-list long boring intellectuals and don't know that you exist. But there is overlap. The forums have the potential to reach critical mass, with enough commentators to sustain interest. And "freedom of speech" benefits from all content being opt-in. Every-one can ban any-one for any-reason, and that ban doesn't extend beyond their personal version of the forum.
More options
Context Copy link