CS (frmr Stat) em. prof. Awarded
#rstats
author, 2016. New book: Art of ML. Teach., pub. service awards. Frmr Ed-in-Chief, R Journ. 1st-gen college. Views mine.
@stephsvox
@UCLA
This is so extreme that one's first impulse is to wonder if it is fake. But yes, all indications (see yesterday's video) are that, sadly, this is the real thing. Again, as a Jewish LA native and UCLA alumnus, I am absolutely amazed both at the students and the administration.…
@thestustustudio
In which Skadden LLP wakes up to find there are much better candidates at non-Ivy schools. Yet another sad silver lining to the tragedy of the last 4 months.
#rstats
Here I will first introduce the concept of quantile regression (QR), a powerful technique that is rarely taught in stat courses. I’ll give an example from the quantreg package, and then will show how qeML can be used to do model-free QR estimation. 1/
@DrewPavlou
E-mail has a paper trail, from start to finish. Had this been checked, the police would have seen it did not originate with you. Conversely, you could use it now to bring a charge against the embassy of filing a false police report.
#rstats
Yesterday I stated that
@MattDowle
is one of the most important people in the history of R. Regardless of whether you use
#rdatatable
or not, it's important that you understand why. 1/n
@tcarpenter216
The best advice I ever got on writing (actually from a nonacademic): Write a draft; put it aside for couple days; come back then, read it again, and be horrified :-) ; rewrite. Repeat cycle.
@Hebro_Steele
@Stanford
@joboaler
What she advocates amounts to establishing Jim Crow in California schools. The Data Science course she supports in lieu of Algebra II is populated mainly with Black and Latino students.
@thegautamkamath
Good, but in the vast majority of cases, the kids will be "advised" by their engineer parents or other adults, even researchers in the field.
@DrJBhattacharya
Even Teddy Roosevelt's "bully pulput" was out in the open, as was JFK's pressuring of the steel makers. To compare those with the feds' clandestine pressuring of the social media firms is absurd.
#rstats
It's been great that so many people have commented on my "R vs. Python" essay, . I've made various improvements, and will continue to do so as more comments come in and/or I have some new points to add. 1/2
I've set up the Web page for my new book, The Art of Machine Learning, . There are resources there, e.g. code, and an explanation of the book's philosophy.
@ryxcommar
If it is true, I believe one of the reasons may be that employers have figured out that their existing staff can quickly learn DS through self study, no need for specialists.
@ProfJanelleWong
Highly misleading claim. Look at the data, which show that the perpetrators are DISPROPORTIONATELY black. Note too that media coverage tends to focus on the violent crimes, as they should. A white teenage who speaks an epithet is hurtful, yes, but not like physical assault. 1/
@realchrisrufo
Virtually everyone in China, even in the hinterlands, knows that Americans elected a Black president, and now a Black vice president. Most know who Steph Curry and Lebron James are, etc. This video should create a bit of cognitive dissonance, I'd say.
@michelletandler
"Male with machete is back." What a quote!
Friday's WSJ said a downtown office bldg. previously worth $700 now is expected to sell for $60 million.
But no one seems to be able to connect the dots.
@johnmyleswhite
The salient advantage of academia is tenure. When those grad students and post docs go for interviews at a company, they should note what proportion of the workers there are in their 40s or 50s. If it's small, they should ask where THEY will be in their 40s and 50s.
@yudapearl
I believe the main point of the paper was that stat people assume the data come from a population, while for ML people, the data are the data, period. The ML people, thus less constrained, were happy to tackle hard applications that the stat people wouldn't touch. 1/
I've completed a planned update of my fastStat introduction to statistics, and inspired by a Twitter thread today, have added a short section on data privacy.
@f2harrell
has come out with a radically new R tutorial, rflow The name is short for "R workflow," but I call it "R in a box," everything one needs to begin serious usage of R, starting from 0 background. 1/
This is fascinating. I would have thought he'd put down the "mongrel" nature of English (German + French) as "inelegant." But no, to him it's a feature, not a bug. And he likes the prepositions! Good for him--I've always liked to mull break up, break down, break in, break out and…
In 1977, Jorge Luis Borges, one of the most influential 20th c. Spanish-language writers, told William F. Buckley (whose first language was Spanish) his reasons for feeling, age 78, that English was 'far finer' than his native tongue. Right or wrong, I love his savor for language
#rstats
I think R beginners should get a good grounding in base-R, whether they go on from there to Tidy, data.table or neither, I fear that many are being taught only the fancy stuff, rendering them helpless on simple, commonplace tasks. For example:
@KatyaSedgwick
The victim died in trying to fight back for the goods robbed from her. If she felt so strongly about social inequity, wouldn't she have just let them take the goods?
#rstats
, on the importance of using a debugging tool: This is a lifelong major issue with me, and I actually wrote a book on it some years ago. If you write R code of more than a few lines, don't use print() as your debugging device! 1/
Each year around this time, I update , my advice to undergrad students in Computer Science, Data Science and Statistics, esp. first-year students. ("What no one else will tell you")
@groundwalkergmb
@hardsci
@deaneckles
But that is exactly
@hardsci
's point, Gabe. The fact that the GRE doesn't matter much among those who are admitted to grad school doesn't mean that the GRE is irrelevant as one of the factors in admission.
@PhDemetri
To me, all the really important properties of the linear model are distribution-free. One doesn't need conditional distributions or matrix algebra to teach it either. Writing something like
mean weight = a + b height + c age
is good enough. 1/
The exchange I had the other day with 2 tidyverse supporters here inspired me to collect some data re usage of base-R vs. Tidy in major R packages. I've just posted it to my stat/R blog,
@EconTalker
The key, usually not considered, is as you say, "They've promised to do this again and again." I would add that their technological sophistication and boldness is apparently much keener than in the past. 31K Palestinian deaths is horrific, but Hamas now seems to have the means to…
#rstats
for YOU: Ever encountered an R error message like "Couldn't create memory segment of size 3.2G"? Also, more subtly, ever had some code run slowly in spite of little apparent reason? And ever wonder why data.table is so blindingly fast? This post will be on MEMORY. 1/n
#rstats
Recently a young relative mentioned that the campus R course she hoped to attend was full. What online alternatives did she have? So, I decided to start one of my own! Designed for complete beginners. Have 5 lessons so far.
The late Leo Breiman is usually credited with inventing random forests, but actually the idea was proposed earlier by Tin Kam Ho, who called it "random decision forests." Sadly, she is seldom mentioned in lists of top stat/machine learning people, despite her major contribution.
#rstats
for YOU: Did you know that R's Principle Components Analysis workhorse, prcomp(), is paired with a predict() function? In this post, I'll explain why that is useful. If you don't know PCA or need review, I'll fill you in. 1/n
In each course I teach, I assign a term project, in groups of 3-4. The topic this quarter was to choose a CRAN package and improve its speed by converting parts to C/C++. Several were really excellent; I've chosen this one to display (with permission),
@thepointwithlx
Cantonese IS Chinese, and is what most HKers prefer. But most HKers can understand Mandarin today. On the other hand, there is so much resentment against inland China that many will not respond to Mandarin. Those from inland should ask themselves why.
I've been writing about this for many years. The industry has always claimed a "shortage" of software engineers, claiming the solution is to get K-12 schools to improve instruction in STEM--at the same time the industry is laying off older engineers who excelled in STEM in…
My dad has run into this many times. He built two supercomputers (ASCII red + blue), has code in the Linux kernel, built a novel operating system, is a top GPU person, etc. and has been turned down by Nvidia/Google when he had offers for them many years ago
@dabeard
@JiayangFan
She was the first writer in the mainstream press to expose abuse of the H-1B work visa, long before bipartisan recognition of the problems. See this article quoting
@EricRWeinstein
@alexandrosM
This long thread is not for everyone. But since Malone's name is coming up increasingly often, esp. in political circles, the thread is worth a read.
@EricRWeinstein
Some years ago, some grad students sued UC, claiming they were workers. Then someone from my campus administration ordered faculty to disagree. Yes, you read this right. Of course, we just ignored them.
@Coqui6190
@noahlet
@beatrizurdag
I still think fondly of my first visit to Hong Kong. Though it was a personal trip, I dropped by the University of HK Stat Dept., no advance notice at all. The dept. chair took me to lunch, and invited me to give a talk. The CUHK saw it, and then invited me too.
@rastokke
@FightFuzzyMath
Any creative mathematician has automaticity with math facts. It forms a vital base for the creative work. Can't be creative if you have to constantly be distracted by having to look up 3 x 8.
@kareem_carr
Actually, the claim is accurate and relevant, and your explanation confirms it. If inequities are to be addressed, the root causes must be identified. E.g. if, given a good level of education, hiring is fair, then this identifies the problem as education, not hiring. Then one…
@JakeVigdor
It cites evidence found by the universities themselves.
IMO this is an excellent piece, with an important subtheme that there was a major Bandwagon Effect underlying the original decision to drop the test.
@yuanyi_z
"We should teach them statistics" has become the standard dog whistle for "We should water math curriculum down." REAL stat would be great, even vital, but it requires familiarity with functions, slopes of lines etc.--which the "reformers" want to omit from curricula.
@PhDemetri
Students come away thinking normally distr. epsilon is crucial. It isn't. For large n we have approximate normality of beta hat anyway, while for small n the "exact" methods make an assumption that is never true in real life. Students deserve better than what we are giving them.
@Twolfrecovery
What stands out in my mind was her dismissing crimes such as car break-ins by saying, "Look, nobody got hurt." On the contrary, it is quite traumatic, especially to kids, to find someone has smashed one's car windows and stolen things.
@MaartenvSmeden
When we released , which argues that neural networks are actually old-fashioned polynomial regression, quite a controversy ensued. The ML people were defensive (to the point of even criticizing the fonts in our paper), but the stat people were happy. :-)
#rstats
for YOU: As R surged in popularity the last few years, the skeptics developed some slogans, one of which is "Yes, R is fine for informal usage on small tasks, but it's not suitable for production," referring to large jobs run on a regular basis, say every hour. 1/n
@dannoyes
"Judge Richard Kramer: 'Of course, I'm frustrated [about rampant, largely unpunished crime in SF]. Why do you think I'm living in Napa now? I lived in San Francisco forever.'"
For the last couple of years, many of us have been concerned with the proposal to allow Data Science as an alternative to Algebra 2 in CA high schools. Many STEM academics and professionals have signed petitions against the plan. 1/
@MaartenvSmeden
Linear models run our lives, pretty well if done carefully. If one adds polynomial terms, say up to degrees 2 or 3, one can often compete well with ML algorithms, while still using lm().
@melissakchan
A more charitable view: The Chinese term for "democracy" can easily be taken as "people-ism," covering at least the "for the people" in Lincoln's famous line. I'd say most Chinese view their gov. in that light (in principle), so it's not just indoctrination.
@daniela_witten
Ironically, I'm currently reading mathematician Shing-Tung Yau's autobiography, and half the book seems to be about his enemies. I know a couple of the people mentioned as non-enemies (as do you :-) ), rather startling reading.
I've just completed fastStat, , a quick intro to statistics for those who've had a calculus-based probability course. Many such people later need to do stat, and this will give them quick access. Modeled after my R tutorial, 1/
The recent article , which compared Python and R, contained a very misleading statement. In explaining that, I can make this post a mini-tutorial about parallel computation. 1/n
My biggest concern, though, is the proposal's methods of instruction. Teachers would de-emphasize getting the right answer, and would not require students to show their work. Emphasize would be placed on collaborative group work, rather than individual efforts. 6/
Jo Boaler speaks out re the charges brought against her .
She says "all research used for the [California Math]Framework was thoroughly vetted by an independent party," and that "most of their accusations demonstrating a lack of understanding of…
@deb_fillman
Excerpt: "I grew up on Obama. Democratic household. I went to his inauguration...[Before my high school teacher told me later on] I had no idea he [Obama] was a terrorist." Mind boggling statement! Well, she's making up for lost time now, I guess. :-( Yes, an Ivy League student.
Recently I speculated that Dr. Gay's years of unethically copying research material could be due to her lack of understanding that material. Here prominent stat prof Andrew Gelman says this too, more forcefully. 1/
@realchrisrufo
I continue to believe that the "grievance industry" people view a PhD as simply a "union card" needed to practice their trade. It's not the classical pursuit of intellectual inquiry. Thus "shortcuts" are not surprising, though still disappointing.
#rstats
Torch for R is now in the qeML package.
I’ve added a new function, qeNeuralTorch, to the qeML package, as an alternative to the package’s qeNeural. It is experimental as this point, but usable and I urge everyone to try it out. In this post, I will (a) state why I felt…
@StanleyRoberts
@FosterCityPD
@MrBadlyTV
While there are legitimate concerns re "driving while black," stopping a car with blacked-out license plates is legal and highly justifued.
@stephaniemlee
"...Youcubed, a Stanford research center that she co-founded to provide free teaching resources": YouCubed is a "revenue generator" for Stanford. They do have some free materials, but they charge fees, e.g. $995 per person for a summer workshop for teachers…
@lpachter
:-) Once at a conference I got to a session early, and sat in the front row. Rather large lecture hall. I saw a janitor enter the room, dressed rather shabbily. Then I noticed that the janitor was walking toward me, getting closer and closer rather than veering off. Then he sat…
@Pal_action
It's an interesting question, whether victors in war have "rights." Questionable, I guess. But the entire history of the Middle East is riddled with victors exerting such "rights." The "owners" of Palestine before the British, the Ottomans, did so, for instance. For that matter,…
I urge CA STEM professors to sign a letter that will be delivered to Gov. Newsom, regardless of where you are on the political spectrum. Only an organized effort can have an impact. 10/10