Top all-time · #311
Andrew Gelman
Personal blog · stat.columbia.edudatascience
Andrew Gelman writes Department of Statistics, a personal blog at stat.columbia.edu about data and science. Since 2008, 163 different Hacker News readers have linked it 218 times, most often “The Garden of Forking Paths [pdf]”. It is still writing: the last post was on 2026-10-21, and you can follow it by feed.
Rarely recommended · 5 of 40 sampled comments recommend it?
Why HN links it
HN readers link Gelman's pages mostly as sources on forking paths, p-values and Bayesian statistics. A few point to his book, papers and Stan primer as good reading.
- Recommended for · 5 comments
- Called 'does a good job', 'a really good read', 'drove home many, many points', 'good textbooks', 'helpful links'
Summarised by Claude from a sample of 40 HN comments that link here. Each comment is marked recommending, critical or neutral (citing without judging); how this works.
Most-linked posts?
- The Garden of Forking Paths [pdf]19 people · 19 mentions · first linked 2015-08-18
- /~gelman/book/13 people · 13 mentions · first linked 2014-08-29
- You Can Load a Die, but You Can’t Bias a Coin (2002) [pdf]12 people · 12 mentions · first linked 2011-10-19
- /~gelman/arm/7 people · 9 mentions · first linked 2009-02-21
- Philosophy and the practice of Bayesian statistics [pdf]6 people · 6 mentions · first linked 2013-03-27
- /~gelman/research/published/kanazawa.pdf3 people · 3 mentions · first linked 2009-07-28
- Why Sell Seashells by the Seashore: Implicit Egotism and Life Decisions (2002) [pdf]3 people · 3 mentions · first linked 2015-01-13
- Healthcare spending and life expectancy: a comparison of graphs3 people · 3 mentions · first linked 2010-01-03
- Statistical challenges and misreadings of literature create unreplicable science [pdf]3 people · 3 mentions · first linked 2024-10-28
- /~gelman/research/published/philosophy_online4.pdf2 people · 2 mentions · first linked 2016-01-24
What Hacker News says?
It's easy to assume the worst of researchers, but I think Gelman's 2013 "Garden of Forking Paths"[1] does a good job of showing how it doesn't have to be intentional misconduct. It more likely or more often comes down to implicitly selecting your analysis plan given the data, and not correcting for the (hidden) multiple comparison problem that arises when doing frequentist statistical tests in that reality: the theo…
Comment linking /~gelman/research/unpublished/p_hacking.pdf
on “Daniel Kahneman “I placed too much faith in underpowered studies””Submitted What are the most important statistical ideas of the past 50 years? [pdf] as a story · 288 points
HN discussionNot my impression. The kind of outlier-culling technique suggested by civilized is not recommended these days because it adds unprincipled choice points to what Andrew Gelman calls the 'Garden of Forking Paths' [1, 2]. Thus they are bad for hypothesis testing, which tends to be what most statisticians care about. Additionally the technique obscures the relationship between the variance of the sample and the populati…
Comment linking /~gelman/research/unpublished/p_hacking.pdf
on “Efficiency is dangerous and slowing down makes life better”Submitted Bayesian stats book banned in China as a story · 159 points
HN discussionThe use of p-values with a research design like you are describing worries me: 'The garden of forking paths: Why multiple comparisons can be a problem, even when there is no \ shing expedition" or "p-hacking" and the research hypothesis was posited ahead of time' http://www.stat.columbia.edu/~gelman/research/unpublished/p_...
Comment linking /~gelman/research/unpublished/p_hacking.pdf
on “Second Gravitational Wave Detected at LIGO”Submitted What made Bell Labs special? (2012) [pdf] as a story · 144 points
HN discussionI made the same transition earlier in my career. One book on deep learning that meets your requirements is [0]. It’s readable, covers a broad set of modern topics, and has pragmatic tips for real use cases. For general machine learning, there are many, many books. A good intro is [1] and a more comprehensive, reference sort of book is [2]. Frankly, by this point, even reading the documentation and user guide of scik…
Comment linking /~gelman/arm/
on “The Matrix Calculus You Need for Deep Learning”Submitted Philosophy and the practice of Bayesian statistics [pdf] as a story · 98 points
HN discussionNot really. Lots of neurons in this part of the brain modulate their activity to movement in heterogeneous ways. The algorithm details vary, but at some level you’re trying to find a 2d x/y velocity signal encoded in the 200d neural signals. This decoder is a bit more sophisticated (using deep learning style approaches), but a Kalman filter was state of the art for a long time. For the adaptation, there’s a rich lit…
Comment linking /~liam/research/pubs/merel-fox-coadaptation.pdf
on “Neural implant lets paralyzed person type by imagining writing”Submitted Statistical challenges and misreadings of literature create unreplicable science [pdf] as a story · 67 points
HN discussionYour ucsb link describes an unfair coin flipping protocol, and not the fair protocol commonly used in practice. In particular, it provides a "strategy" at the end to generate an unfair toss (which requires, e.g., that you be both the flipper and the chooser). Other studies, such as the famous You can load a die, bit you can't bias a coin investigate fair flipping protocols and argue that you cannot, in fact, bias a…
Comment linking /~gelman/research/published/diceRev2.pdf
on “Seeing Theory: A visual introduction to probability and statistics (2017)”Submitted No problem, we'll adjust the data to fit the model as a story · 57 points
HN discussionAt a minimum, recognize that not all p < 0.05 results have an equal claim on scientific truth. The study power and the prior probability of the hypothesis being true also influence the posterior probability [1]. In particular, underpowered studies of implausible hypotheses provide almost no scientific signal [2], and p < 0.05 is meaningless in such cases. [1] https://journals.plos.org/plosmedicine/article?id=10.1371…
Comment linking /~gelman/research/unpublished/power4r.pdf
on “Statistical Rethinking (2022 Edition)”Submitted Fully Bayesian Computing as a story · 53 points
HN discussionThe book is also available on Prof Gelman's department site which I would probably link to instead: http://www.stat.columbia.edu/~gelman/book/
Comment linking /~gelman/book/
on “Bayesian Data Analysis, Third Edition [pdf]”Submitted The commissar for traffic presents the latest five-year plan (2014) [pdf] as a story · 52 points
HN discussionI'm not well rounded enough to draw a clear path from where you are. For me Gelmans Data Analysis Using Regression and Multilevel/Hierarchical Models [0] drove home many, many points. More recently, I have a sense/hope that Pearl's The Book of Why [1] might take this to yet another level. [0] http://www.stat.columbia.edu/~gelman/arm/ [1] https://www.amazon.com/Book-Why-Science-Cause-Effect/dp/0465...
Comment linking /~gelman/arm/
on “Ask HN: What are your favorite statistics and probability textbooks?”Submitted What is a standard error? [pdf] as a story · 45 points
HN discussionI thought this was common enough knowledge not to want citations, but I think you will find these satisfactory. http://www.stat.columbia.edu/~gelman/research/unpublished/sw... https://www.politico.com/magazine/story/2014/01/independent-... https://www.thenation.com/article/what-everyone-gets-wrong-a... The last piece has a short summary of the salient point: > In fact, according to an analysis of voting patterns con…
Comment linking /~gelman/research/unpublished/swing_voters.pdf
on “How Cambridge Analytica’s Facebook targeting model really worked”Submitted Validating Bayesian Inference Algorithms with Simulation-Based Calibration [pdf] as a story · 44 points
HN discussion> a) could not reliably capture a one minute flame graph covering the issue, as the onset of it kept fluctuating; b) capturing a two or three minute flamegraph didn’t help either, and the issue was “drowned” in the normal workload profile. Vadim asked me for help. One interesting thing here is also the statistical nature of this problem: i.e. there are spikes but their duration are not exact or known when to start.…
Comment linking /~liam/teaching/neurostat-spr12/papers/brown-et-al/time-rescaling.pdf
on “Netflix FlameScope”Submitted Skepticism about beautiful people having more daughters. as a story · 43 points
HN discussion> or a p value of something like p = 0.0001 This has been proposed [0], albeit for a threshold of p < 0.005. Here's Andy Gelman and others arguing otherwise [1]. They also got like 800 scientists to sign on to the general idea of no longer using statistical significance at all [2]. [0] https://www.nature.com/articles/s41562-017-0189-z [1] http://www.stat.columbia.edu/~gelman/research/unpublished/ab... [2] https://ww…
Comment linking /~gelman/research/unpublished/abandon.pdf
on “Effect size is significantly more important than statistical significance”Submitted Garden of forking paths: Why multiple comparisons can be a problem (2013) [pdf] as a story · 40 points
HN discussionCome into Bayesian land, the water is fine. The whole NHST edifice starts to seem really shaky once you stop and wonder if "True" and "False" are really the only two possible states of a scientific hypothesis. Andrew Gelman has written about this in many places, e.g. http://www.stat.columbia.edu/~gelman/research/published/aban... .
Comment linking /~gelman/research/published/abandon.pdf
on “Effect size is significantly more important than statistical significance”Submitted You Can Load a Die, but You Can’t Bias a Coin (2002) [pdf] as a story · 34 points
HN discussionYou appear to have your facts wrong. From the study[1]: > Eligible cases were all patients with a glioma or meningioma of the brain diagnosed in the study regions during study periods of 2–4 years between 2000 and 2004. So participants in the study lasted between 2 and 4 years. Since elevated rates of cancer were only found in the group with the highest usage[2], this means people with >=1640 hours of talk time, whi…
Comment linking /~cook/movabletype/archives/2010/05/the_bane_of_man.html
on “Antenna designer writes about the iPhone 4”Submitted Review of RStudio (new IDE for R) as a story · 33 points
HN discussion> Is the scientific method, or more generally modern science and how it is distributed to those outside the field, so poorly designed that not disclosing ties invalidates what was communicated? Basically, yeah. There's the famous replication crisis, and of course p-hacking. There's also the forking garden problem, e.g. in: http://www.stat.columbia.edu/~gelman/research/unpublished/p_... Put it like this: If you're mo…
Comment linking /~gelman/research/unpublished/p_hacking.pdf
on “NYU professors who defended vaping didn't disclose ties to Juul”Submitted Classics of statistics as a story · 32 points
HN discussion
Linked by the same people?
- AAndrew Gelmanandrewgelman.com
- SSimply Statisticssimplystatistics.org
- TTanya Khovanovablog.tanyakhovanova.com
- M
Machine Learning (Theory)hunch.net - R
Richard J. Liptonrjlipton.wordpress.com - J
John D. Cookjohndcook.com - C
Cosma Shalizicscs.umich.edu - D
Department of Mathematics at Columbia Universitymath.columbia.edu