Understanding Uncertainty

For up to date work, see the MU Collective [0]: "We are a cross-institution research lab working at the intersection of information visualization and uncertainty communication. Our mission is to improve both experts' and lay people's abilities to reason about data through visual representations that align with how people think. Topics we like include sampling-oriented uncertainty visualizations, interactive visualization for thinking about priors, multiple views, and Bayesian statistics. MU Collective is directed by Jessica Hullman (Northwestern University) and Matt Kay (University of Michigan)" Hullman is also usefully active on twitter [1]. There's lots going on in this area. The posted article links through to "Understanding Uncertainty", [2], which has a mission of improving the public's understanding of uncertainty but appears to be a lapsed project. One member of the project team is David Spiegelhalter, who is still writing. His book "The Art of Statistics" is an excellent book on the craft of inference. [0] https://mucollective.northwestern.edu/ [1] https://twitter.com/JessicaHullman [2] https://understandinguncertainty.org/

Right, but you can find correlations between being black and committing crime - compiling data and running logistic regressions be damned. These measures they came up with are heuristic proxy measures at the very best, and noise at the worst. The troll hunting algorithm has to face the false positive problem [1], which the paper does not address. My very legitimate content has been censored various places (notably Facebook) because it tripped 'anti-trolling and scam algorithms' but the things I was trying to post were Snowden and Manning, and TPP leaks. [1] http://understandinguncertainty.org/node/238

> GPs will tell you that this is wha they say to most patients. Yes, and when that happens, which it certainly does, it's not science. Are you sure an example from medicine that follows the same pattern as psychology, somehow validates the practice as science for both? With all respect, that's not the most powerful argument. A scientific example from medicine would be -- to choose from hundreds of examples -- Ebola, in which we know exactly what causes it, we can view the causative agent in a microscope, and we know how to treat the cause, not the symptoms. It's the same with hundreds of other conditions. No such conditions or treatments exist in psychology -- the DSM is a list of symptoms with no causes. > Given the evidence for CBT is so strong and the evidence for counseling is so weak it's telling that you recommend "chat's with your aunt" as being an effective treatment - totally misreading the literature. You are mistaken. In fact, a careful reading of the literature leads one to exactly that conclusion (i.e. all therapies are equally effective, therefore there is no basis for preferring CBT over IPT, or a psychotherapist over a sympathetic relative). This is a typical crititism of the literature supporting CBT: http://en.wikipedia.org/wiki/Cognitive_behavioral_therapy#Criticisms Quote: "... the methodological processes used to select the studies in the previously mentioned meta-analysis and the worth of its findings have been called into…

Probability is hard to understand and it means different things in different contexts. (is it only about knowledge or is it out there in the world) If it is hard for Andrew Gelman it may be too hard for most of us. "Court of Appeal bans Bayesian probability (and Sherlock Holmes)" https://understandinguncertainty.org/court-appeal-bans-bayesian-probab…

Re-reading, sorry. So ... "probability of surviving another 10 years" is different from life-expectency. It's a variant on mortality, though expressed over a 10 year period rather than a 1 year period. For comparison, see the SSA's actuarial life tables. This gives 1-year mortality (both as percentage and as expected survivors per 100,000 lives) for each year of life from 0 to 119: https://www.ssa.gov/oact/STATS/table4c6.html I'm not aware of a compilation or chart showing changes in actuarial mortality by age over time. There's a good chart of anticipated deaths per year (separately for women and men) in the UK: https://understandinguncertainty.org/why-life-expectancy-misleading-su… What question are you trying to answer by looking at mortality probability?

Dr David Spiegelhalter runs the Winton programme for the public understanding of risk at Cambridge. He has some good stuff on this for more everyday activities like drinking beer, being 5kg overweight for a day and smoking a cigarette and things. https://understandinguncertainty.org/microlives

"In summary, the study is grossly underpowered to convincingly prove a plausible protection, and they have committed the cardinal sin of saying that non-significance is the same as 'no effect' in a study lacking sufficient events, in this case, deaths in non-drinkers. Maybe epidemiological studies should include power calculations, which make sure there is a reasonable chance of detecting a plausible effect, and which became standard in clinical trials after too-small studies were being used to claim that drugs did not work." http://understandinguncertainty.org/misleading-conclusions-alcohol-pro…

Logic is not required of argument in court. In fact certain types of logic, like bayesian reasoning, are sometimes explicitly excluded. http://understandinguncertainty.org/court-appeal-bans-bayesian-probabi…

Here's a peer review: http://michaelgrayer.posterous.com/in-which-i-blow-a-gasket-and-get-ve… Since PRNewswire/Sustainable Food Trust didn't seem to employ a critical eye to examine the conclusions of the study. edit Further criticism of the design of the experiment and the statistics: http://understandinguncertainty.org/rats-and-gm

> Blood/urine test =/= testing. Also from the link you provided: "Screening is the process of identifying individuals who may be at higher risk of a disease or condition amongst large populations of healthy people." People who are not at risk might want to do blood/urine test to check if they currently have underlying issues. This sentence doesn't make much sense. > Blood/urine test =/= testing. If the patient has symptoms then it's a test. If the patient does not have symptoms then it's screening. > People who are not at risk might want to do blood/urine test to check if they currently have underlying issues. But this is screening, and screening is often harmful, which is why it's strictly controlled. The blog is not only about covid-19, it uses Covid-19 testing as an example. But the information about screening, especially false positives, applies to the type of blood testing you're talking about. If you want a different blog: https://understandinguncertainty.org/node/1279 > Last time I checked that was the thing suggested, can you please clarify and elaborate why you say so? The problem is of "over testing", "over diagnosis", and "over treatment". https://ebm.bmj.com/content/23/1/1 Your claim is that early detection means that treatment can start earlier, and that this reduces side effects and prolongs life. My claim is that often screening sometimes does not prolong life, and sometimes causes…