Student: I have no idea how to do this problem.
Me: Did you try some simple and extreme examples?
Student: No, I don't have time to try that.
Me: You don't have time NOT to try that.
YAY!!! So much joy and relief after 4 nightmarish years of a racist, sexist, xenophobic, anti-science, authoritarian, mendacious, narcissistic, corrupt sociopath as president. A huge amount needs to be fixed and improved, but at least there is a restoration of decency.
I'm grateful that so many students want to learn probability with me (Stat 110 is undergrad level, 210 is grad level)! But I'm also relieved to have fallen short of 900 students total, mindful of the old saying, "When 900 students you reach, look as good, you will not."
I support
@daniela_witten
that
@AmstatNews
Fisher Lecture be renamed. Fisher was a virulent racist and eugenicist. Hard in this context to fully separate his scientific ideas from his personal ideas; he was even a PROFESSOR OF EUGENICS! Shouldn’t be honored as a hero of the field
Thank you Batman for giving a guest lecture in Stat 110 today! (He was returning a favor since I helped him solve some treacherous probability puzzles that the Riddler had posed him.)
by
@SherlockpHolmes
and
@wolfgangkhuber
is one of the most beautiful and useful statistics books I've ever seen. And I don't even work on biology (thought I want to learn more).
Conflict of interest disclosure: Susan was one of my mentors at Stanford.
Looking forward to teaching statistical inference with Susan Murphy this spring! But there's no book we're excited about using. Too many are full of context-free problems like "Consider the model f_theta(y) = blah. Find the MLE of theta." Do you have cool examples to suggest?
I had 92 students attend my Zoom office hours tonight... definitely one advantage of online, considering that I can barely fit 6 students in my physical office, even without social distancing!
Last night I dreamed that I got to meet
@Lin_Manuel
, and we were excited to collaborate on a musical about the life of a statistician. First sweet dream I can remember having in this nightmarish year. Hoping so much for some good news in 3 days...
Got a bit too excited in class explaining an example with P(A|B,C) where A is the event that there is a fire, until I realized that shouting “fire” in a crowded lecture hall is not the best idea.
I taught John Snow's work on cholera in Stat 111 a few days ago, to show the importance of data quality and causal inference. Couldn't resist saying that people around him thought he was crazy for thinking cholera was waterborne, that they said "You know nothing, John Snow."
Big picture for my course with Susan Murphy and (when not on sabbatical)
@shephard_neil
. Prediction and causal inference are center stage, along with classical statistical inference. Probability, randomization, computing, and Bayesian and frequentist thinking come together!
This is bad. ICE just told students here on student visas that if their school is going online-only this fall, the students must depart the United States and cannot remain through the fall semester.
I read the horrifying 65 page filing
Brave, credible, and devastatingly important — and shows utterly vile conduct by Comaroff for years, and systemically broken investigatory processes at Harvard (slow, incompetent, and failing to protect victims).
Video of the Pickard Lecture by
@minebocek
is now online! A wonderful talk, with lots of food for thought about course design and teaching statistics and data science.
Statisticians are called weird for terms like Law of the Unconscious Statistician (LOTUS), but I just heard that actuaries use "Darth Vader Rule" for the fact that the expectation of a nonnegative r.v. is the integral of its survival function.
Awwww, thanks so much
@zacharylipton
! I'm not sure it is worth the risk to the space-time continuum, but happy to trade with you some probability and inference knowledge for some help with understanding deep learning. :)
I want to go back in time to take undergrad probability & statistical inference from
@stat110
.
Check out his wildly popular videos on probability:
& his charismatic textbook:
Trump at debate in 2016, on whether he would accept result of the election: "I'll keep you in suspense."
Trump today: won't accept result of the election, says to the domestic terrorists he incited "We love you. You're very special."
But keep in mind Clinton used private email.
I'm not a legal expert... can Biden simply "hereby claim" Michigan to counter Trump's "hereby claim"? If both assert "hereby claims", is there some system in place for determining who gets Michigan's electoral votes?
@economeager
Yep, they seem like they should be mostly harmless but they're very treacherous! Check out Larry Wasserman's post "Mixture Models: The Twilight Zone of Statistics" (great title!)
and this paper
by
@AviFeller
, Greif, Miratrix, Pillai
The entire Statland saga in one handy playlist:
story proofs, P(A|B) vs. P(B|A), discrete distributions, interplay between discrete & continuous, expectation and linearity, conditional expectation, and Markov chains, animated in about 5 min each!
#stat110x
Congratulations
@BaumerBen
on the 2019 Outstanding Undergraduate Teaching Award from the Boston Chapter of the ASA! He is giving a talk "The New Roaring Twenties: Imagining statistics and data science curricula in the coming decade" at Harvard on Oct 7:
Charles Murray promotes racist pseudoscience, not data science. For data science it is essential to think about data quality, potential biases in the data, and data ethics.
I always liked Mosteller's quote "It is easy to lie with statistics; it is easier to lie without statistics." But Trump's thousands of lies give many counterexamples: he lies with equal ease both with statistics and without statistics.
@Harvard
needs to allow a 3rd party grievance procedure and to completely rethink how such cases are handled (and not just create some new committee that will spend years debating what went wrong here and then issue a report that will mostly be ignored).
@willemsleegers
Which do you care about more, sample variance or sample SD? As soon as you take the sqrt of the unbiased sample var it will be biased anyway (by Jensen). Also, for Normal data, dividing by n-1 is unbiased, dividing by n is the MLE, but dividing by n+1 is optimal (w.r.t. MSE)!
Visual interpretation of Universality of the Uniform (a.k.a. the Probability Integral Transform) by
@brianczhang
:
Brian made this 5 years ago as a Stat 110 TA, but students are still finding it useful!
@StuartBuck1
They call it an "example", but it is a horrible example. And doesn't make sense, how are the terms partitioned up? I was surprised to see data science on the "easy to learn" side, but guess that assumes the "hard to learn" subjects of math, statistics, ML are excluded from DS...
I enjoyed
#JSM2010
in Vancouver, but made the mistake of almost never leaving the convention center except to go to the hotel. Rectifying that at
#JSM2018
. Just went to Capilano Suspension Bridge Park, which was amazing! Highly recommended (unless you’re very afraid of heights)
@Notawful
I think it's just because a hundred years ago Student (Gosset) and Fisher happened to use the notation z for what is now called a z statistic, and likewise for t and the t statistic. Statisticians are usually not great at naming things (the bootstrap is a notable counterexample).
Looking forward to the David K. Pickard Memorial Lecture by
@minebocek
this Friday, on data science course design and pedagogy! Title, abstract, and more information at:
If you are in the Boston area, you are welcome to attend!
RIP to my friend and
@Harvard
colleague Prof. Rob Lue, a visionary not only in biology education but also in the interface of science & art in his leadership at
@DerekBokCenter
@HarvardOnline
@TheHDSR
& more. Check out his Inner Life of the Cell animation
He needs to be fired, for the massive harm he has done, to prevent him from doing more such harm, and to avoid sending the message that such conduct is acceptable. Tenure is meant to protect academic freedom, not to protect harassment and abuse.
Welcome
@shephard_neil
to Twitter! Neil is Harvard's first ever joint Econ-Stat professor, a great financial econometrician, and has had a huge impact on statistics and data science at Harvard since arriving from Oxford 5 years ago.
I enjoy the many insightful, funny, and insightful-and-funny tweets of
@kareem_carr
. Let’s support thoughtful, respectful discussion — terrible that there are so many racists, trolls, and racist trolls attacking him.
Thought it might be useful to collect screenshots of people who brought my race into the discussion of my thread even though I’m just talking about arithmetic. It helps to bring context to the underlying reasons for the objections to what I’m saying and who’s saying it.
My favorite musical is Into the Woods. One of my favorite probability puzzles is the Monty Hall problem. Today I learned that Joanna Gleason, who played the Baker's Wife in the original Broadway production, is Monty Hall's daughter.
@hadleywickham
Congratulations Hadley! Extremely well-deserved, and a sign that the field is moving in the direction of "impact t he world by being useful" that you wrote about at
@matt_blackwell
iPad + Apple Pencil + Notability, with interaction from chat and breakout rooms: good
Blazing through hundreds of PowerPoint slides on Zoom: bad
The Efron-Morris baseball example of Stein's paradox and shrinkage, as in their wonderful Scientific American paper
Beautiful, surprising, and useful at the same time.
This is the email I received from Gary Urton. I was a second to last year graduate student and he was chair at the time. Harvard, this time please don’t say there “ was no evidence for violation of policy”.
Trump tonight: “When you do testing to that extent, you are gonna find more people, you’re gonna find more cases. So I said to my people: ‘Slow the testing down, please.’”
Health officials say this is the exact opposite of what the US should be doing right now.
@ProbFact
TWO proofs of this in one tweet:
1. MGF of Pois(a) plus indep Pois (b) is product of MGFs:
exp(a(e^t-1))*exp(b(e^t-1))=exp((a+b)(e^t-1)), the Pois(a+b) MGF.
2. In a Poisson process of rate 1,
# arrivals in (0,a+b) = # in (0,a) + # in [a,a+b)
and # in (0,t) is Pois(t) for all t.
Thank you
@KLdivergence
for your brave, important, and deeply disturbing post at . Horrific and shameful behavior, and the culture needs to change to prevent this from continuing to happen.
@shephard_neil
Until a few years ago, his office was across from mine and he came to the office almost every day! He even asked me questions about
#rstats
, in his 90s, since he was coding up some simulations.
@kbarley66
The Cauchy-Schwarz Master Class (Steele), Applied Linear Algebra (Boyd and Vandenberghe) - free via , Enumerative Combinatorics (Stanley) - free version of Vol. 1 via , Concrete Mathematics (Graham, Knuth, Patashnik)
CDC Director: It would be “very disappointing” if schools use CDC reopening guidelines as guidelines for reopening. I mean, what did you think those reopening guidelines were? Guidelines for reopening or something?
Short Course on the R package Shiny
Instructor: Mine Çetinkaya-Rundel, Senior Lecturer, School of Mathematics, University of Edinburgh
Date and Time: October 16th and October 23rd, 1-4pm
Location: Virtual
Registration:
I've been accused of making up too many acronyms, but differentiation under the integral sign is so cool and useful that it really deserves one. So here it is: DUThIS (Differentiation Under The Integral Sign).
Great materials from a survival analysis workshop by Dave Harrington: …
Data sets that go along with the lectures and labs:
…
#rstats
#survivalanalysis
@heyitsmehugo
@minebocek
@BecomingDataSci
Check out Berger, "Could Fisher, Jeffreys and Neyman Have Agreed on Testing?"
and Christensen, "Testing Fisher, Neyman, Pearson, and Bayes"
So yesterday I asked you all what you wanted to hear about from me this week, and one answer stood out from all the others: the SINGULAR VALUE DECOMPOSITION (SVD).
1/