Francis DiTraglia
@economictricks
Followers
1,137
Following
270
Media
16
Statuses
172
Applied Econometrician, Associate Prof. of Econ @OxfordEconDept and @lmhoxford , visiting scholar Chicago FRB
Oxford UK
Joined April 2020
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
Kendrick
• 251220 Tweets
Defense
• 210161 Tweets
Drake
• 133075 Tweets
Patriots
• 132264 Tweets
Bills
• 124045 Tweets
Giants
• 109185 Tweets
Bears
• 96188 Tweets
Steelers
• 90654 Tweets
Tyreek Hill
• 88919 Tweets
Wayne
• 86170 Tweets
Bengals
• 71070 Tweets
Saints
• 68479 Tweets
Panthers
• 58052 Tweets
Dolphins
• 56849 Tweets
Cowboys
• 56169 Tweets
Vikings
• 53133 Tweets
Falcons
• 49701 Tweets
Lamine
• 48704 Tweets
Fields
• 48495 Tweets
Titans
• 47760 Tweets
Colts
• 40283 Tweets
Texans
• 36274 Tweets
Super Bowl
• 36093 Tweets
Cardinals
• 29079 Tweets
Daniel Jones
• 25037 Tweets
Fritz
• 21363 Tweets
Caleb Williams
• 20523 Tweets
Anthony Richardson
• 20281 Tweets
#HereWeGo
• 19588 Tweets
Josh Allen
• 19102 Tweets
Bryce Young
• 16278 Tweets
TJ Watt
• 13872 Tweets
Zalewski
• 13052 Tweets
Pats
• 12769 Tweets
Sam Darnold
• 12655 Tweets
Jags
• 11119 Tweets
Tom Brady
• 10343 Tweets
Pinned Tweet
📣 Announcing SQARE FAQ! We've compiled answers to your most pressing questions about our proposed quality assessment system for econometrics research. Visit for the full details.
#EconTwitter
#AcademicPublishing
#Econometrics
@WeidnerEcon
1
4
8
Open science -> open source. Still doing your econometrics in STATA or Matlab? Here's a tweetstorm of my favorite free resources to help you make the transition to R. (Julia and Python are also great - I just don't know them as well)
#EconTwitter
#rstats
5
169
750
This flowchart from "Bernoulli's Fallacy" by
@aubreyclayton
made me chuckle over my coffee this morning:
6
57
244
I'm coming around to the idea that we need to teach DAGs. The textbook advice is to control for anything that's correlated with both treatment and outcome without being caused by the treatment. This is wrong, but it's very hard to explain why verbally.
@causalinf
#EconTwitter
2
18
125
Happy Friday
#Econtwitter
! To make sure you have something meaty to read this weekend, here's my longest post yet at , a deep dive into regression adjustment for causal inference
#Econometrics
2
29
104
Hey
#EconTwitter
: ever wondered by the James-Stein estimator takes the form that it does? My new post answers all of your burning questions with nothing more than first-year calculus and introductory statistics!
#Econometrics
2
13
68
Excited to learn that I've been awarded a Public Policy Challenge Fund grant for work with
@LudoGazze
on childhood lead exposure in England. Huge thanks to
@abicadams
for encouraging us to apply and Charlotte and Irini for amazing admin support!
@OxfordEconDept
@lmhoxford
0
1
51
Learn programming basics in R, with an emphasis on simulation experiments and vectorization:
1
3
40
Perk of a career in
#Econometrics
: control over chatty seatmates on flights.
"What do you do?"
"I'm an economist" → [Barrage of questions about inflation/UBI/Bitcoin]
"I'm a statistician" → [Awkward sympathy face, renewed interest in safety card]
#econtwitter
3
0
36
Why not replicate Mostly Harmless Econometrics in STATA and R and Python & Julia?!
2
3
35
Having a blast this week teaching Treatment Effects at
#OxEconSummerSchool
this week. Today we discussed testing the LATE assumptions; tomorrow we'll start looking at Marginal Treatment Effects. Lots of requests for more on Diff-in-diff so I know what to add for next year!
2
2
33
Want to become the world's best research assistant? All you have to do is read these three books:
1
4
27
Consider Figure 7.4 from Hernan & Robins' "Causal Inference: What If?"
We want to know the effect of A on Y. L is correlated with A and with Y and isn't caused by A. Nonetheless, it's a "bad control."
2
7
25
Excited and grateful 🙏 that made it into
@tylercowen
's assorted links this past weekend. As a daily reader of
@MargRev
for nearly 20 years, it means a lot and definitely encourages me to keep writing!
#Econometrics
#econtwitter
0
4
26
I've programming in R for 16 years and, much to my embarrassment, only just realized why and when to use sweep(). What other crucial base R functions have I failed to appreciate?
#RStats
5
2
24
Imbens or Angrist or Heckman might say that we don't need the diagram to realize that L is a bad control. I freely admit that it wasn't obvious to me until I saw the diagram! Analogy: we use algebra and a place-value number system to teach things the Greeks derived verbally.
2
4
23
Some R basics (using tidyverse tools) aimed at STATA users:
1
2
22
Know which econometric technique you want to apply, but don't know which R package to use? Check out the CRAN Task View for econometrics:
1
1
22
Yesterday
@WeidnerEcon
and I unveiled , an initiative for changing the way we evaluate research in
#Econometrics
.
Today, with 47 supporter signatures under our belt, I'd like to answer a simple question: why do we need ?
(1/5)
1
5
21
Here's a
#Bayes
Theorem puzzle for you
#Econometrics
and
#econtwitter
folks: would you rather improve the sensitivity or specificity of your diagnostic test?
1
6
20
A bilingual dictionary to help you translate your existing Matlab code into R:
1
1
20
📊 Summer of Bayes Zoom Reading Group
🔍 Macro/Time Series
📅 2024-07-23
🕑2pm (UK)
🌟 Special Guest: Frank Schorfheide
📚 Sims & Uhlig (1991), Giannone et al (2015, 2019)
🎓 All are welcome!
📧 To attend: bayes [at]
#Econometrics
#EconTwitter
#Bayes
2
4
18
Think the academic journal system is outdated? We do too. Let’s modernize it with ! Join our initiative in
#Econometrics
by signing the support form. First, we revolutionize Econometrics; then, we tackle the rest.
#ResearchInnovation
@WeidnerEcon
0
4
14
Thanks to everyone who participated in my Treatment Effects course at the
#OxEconSummerSchool
and to Emma and Eli for fantastic organizational and logistical support! It was a great week and I look forward to next year :)
1
1
12
@OxfordEconDept
Looking forward to the summer school! It was a blast last year and I still have three weeks to come up with even more cringeworthy jokes about causal inference than I made in the video ;)
Have you read 'Mostly Harmless Econometrics'?
Wondering what's next?
Join Frank DiTraglia for his summer school on Treatment Effects this September!
Register now for both in person and online places:
0
16
72
0
1
12
To make this concrete:
A = physical activity
Y = cervical cancer
L = Pap smear result
U1 = pre-cancer lesion
U2 = health-consciousness
2
1
9
@OxfordEconDept
Want to learn some
#Econometrics
at Oxford this September with a dash of
#Rstats
on the side? Why not sign up for a double-header with me in the morning and
@maxkasy
in the afternoon!
0
4
10
Three weeks into my
#RStats
based empirical research methods course and having a blast! You can follow along here if interested: (pull requests for typo corrections on github are greatly appreciated!)
0
5
9
Somehow I missed this: GitHub copilot integration is available in RStudio version 2023.12.1
I've got the feeling I'm about to become much more productive but marginally worse at R programming...
#rstats
1
4
9
The FWL theorem is crucial, but cumbersome to say aloud, even as an acronym. Now that we know Yule discovered it first, it's getting even worse: Yule-Frisch-Waugh-Lovell!
A modest proposal: let's call it "Yule's Rule" - short, catchy, and historically accurate.
#econometrics
0
0
9
@OxfordEconDept
You've read "Mostly Harmless Econometrics" so what's next? Find out in my September Summer School Course "Treatment Effects: Beyond the Basics"
0
1
8
In response to another question, here's a paper I've been meaning to read for a while on the distinction between anticipation effects and endogeneity:
So many papers; so little time!
0
0
7
@OxfordEconDept
Spending my afternoon designing a course logo definitely doesn't count as procrastination, right?
1
0
5
🚀Ready to level up your
#Econometrics
game? 📚 From spillovers to MTEs, my September course on treatment effects goes beyond the basics.
🎥Promo Video:
🎟️ Register:
🌐 Website:
@OxfordEconDept
#econtwitter
0
2
6
May 16th 2024: launch of
June 17th 2024: AEA/EEA/ES/RES release "survey to help improve the publication process."
Coincidence? 😉
@WeidnerEcon
1
0
6
In response to a question from class today: here's a paper I like that suggests a random coefficients approach for settings with a binary instrument and continuous treatment.
0
0
6
Now that researchers at the USDA Forest Service are interested in machine learning, there's scope for some truly amusing abstracts:
0
0
6
I'm using the discussion board for my empirical research methods course and it's amazing: full LaTeX math and markdown support along with the ability to execute code chunks in R, python, and many others. It's even GDPR compliant so you won't get sued!
1
0
5
If you have any favorite resources that I haven't covered I'd love to hear about them!
7
0
5
Who wants to join me for a virtual reading group on Bayesian methods this summer?
#SummerOfBayes2024
#Econometrics
#Bayes
0
1
5
@SophieStats10
This is such a great thing to be doing: every time a government analyst switches from Excel to R an angel gets its wings :)
0
1
5
@Josh_Merfeld
@WeidnerEcon
2-stage evaluation helps. Stage 1 is more about correctness than "importance" so should be less subjective. Stage 2 begins with papers that have already passed a quality hurdle and only worries about importance relevance. (2/3)
1
0
4
@lavaredmonds
Also helpful if you need to raise your monitor a few inches: I have definitely used MWG for that on multiple occasions...
0
0
4
Thanks to
@d_spiegel
for a great talk at
@OxUniMaths
yesterday and for answering my question about how to improve university-level stats teaching. I was unusually excited to be teaching logistic regression this morning!
1
0
4
I'm designing a 1st-year master's level course called Core Empirical Research Methods
@OxfordEconDept
that aims to teach methods and skills that are crucial for applied research but get left out of most econometrics courses. I have lots of ideas. What are yours?
#EconTwitter
1
2
4
Not the second-hand econometrics textbook I was expecting to receive (small print at top), but if this turns out to have been secretly ghostwritten by Maddala I will totally read it.
#MetricsToTheFace
1
0
4
@ajordannafa
Just like "as n goes to infinity this test statistic approaches a standard normal distribution" 😜
0
0
3
@malleshpai
Preach brother. And profs: never even hint that you think a question is "stupid." One of my grad school classmates was castigated for asking a question about the MV normal distribution: "you should know that!" We stopped asking questions in that class and learned very little.
0
0
3
@Josh_Merfeld
@WeidnerEcon
There's also a lot of mechanisms for combining rater information that look nothing like the status quo and could be worth experimenting with. There's been very little exploration of the space of possible ways to design our evaluation process. (3/3)
0
0
3
@Taylor_Z_Lange
@rlmcelreath
Jumping on the Hacking Bandwagon here, I often recommend "An Introduction to Probability and Inductive Logic" to my students
0
0
3
April fool's day is almost upon us, so here's my attempt at econometrics humor from a few years back:
#Econometrics
1
0
2
@camjpatrick
@john_t_ormerod
@ChelseaParlett
I broadly agree with all this. What I would say is that Bayes + model checking provides a nice way of playing around with additional information that might be hard to incorporate otherwise, a point I tried to make here:
1
0
2
@pedrohcgs
@dlmillimet
I read Leamer's "Specification Searches" during lockdown and liked his perspective on this: if applied people are playing with the toys we've given them in weird ways, maybe we need to rethink our approach to toy design and give them something closer to what they actually want.
1
0
1
Reason
#1
: The function of journals has changed.
In the early days, a key role of ReStud / ECMA / JoE etc. was disseminating research.
Today handles dissemination just fine and the key role of journals is peer review and quality evaluation.
(2/5)
1
0
2
#SummerOfBayes2024
is happening! 1st meeting: Tuesday July 9th @ 2pm UK time on zoom. 2nd meeting 2 weeks later.
Got an email address ending in ".ox.ac.uk"? Register here:
Otherwise email me; all are welcome!
#Bayes
#Econometrics
1
0
2
@thesamasher
Swirl is a good option if you're using R:
I also have some related teaching material that might be helpful, mostly tutorials for students to work through on their own:
1
0
2
It will be interesting to see if this can get off the ground. Let's give it a try and see!
@DinaPomeranz
If you want to join the server,
1. go to
2. create an account on the lefthand side
3. confirm the email
4. Done!
18
95
363
1
0
2
@SpinningHugo
@rmappleby
Agree: build build build esp in key part of the SE. On the regional dimension, have you seen this from
@annastansbury
and co-authors?
0
0
2
@jenniferdoleac
Research departments at the Fed may be a good comparison case: no tenure but relatively good job security and no chilling effect on research that I can discern. My impression is that you would totally be fired if you didn't actually produce the labor memo!
2
0
2
@jenniferdoleac
Senior professors who have stopped doing research and phone in all their teaching, on the other hand, are safe in academia. This definitely seems like a problem we need to think harder about.
0
0
1
@mikedecr
Very cool! My one foray into this kind of work resulted in the following:
I don't think we quite pitched it in the right way, given how long it took to find a journal interested in publishing it, but I still really enjoyed working on it!
0
0
1
I found these maps helpful for thinking about the impact of restricted access to Russian airspace. Seems to make a big difference for London -> Tokyo / Hong Kong. (Also for NY to the same destinations.)
1
0
2
A new post on to whet your appetite for the upcoming virtual Bayesian reading group session on macro / time series econometrics:
#Econometrics
#Bayes
0
1
2
Sorry
@UniofOxford
, but this is wildly misleading. Students who choose to reply to a vaccination survey are likely very different from those who choose not to reply. If your estimate is seriously biased, constructing a confidence interval is pointless.
0
0
2
@JoachimSchork
Nice visual of the textbook ideal conditions for ordinary least squares. But whether to consider these assumptions of linear regression depends on what theorem you're trying to prove. Worth emphasizing that most of these can be dropped depending on what you're trying to do.
1
0
2
My co-authors and I recently revised our paper on spillovers and non-compliance:
Our motivation for this project was Crepon et al. (2013) who study the direct and indirect causal effects of a program that provided job placement services in France. (1/15)
1
0
2
#EconTwitter
, what dataset would you use as a playground for teaching Economics grad students about data cleaning and merging?
1
1
2
@LudoGazze
This came up for me yesterday: I clicked "accept" and added a message explaining the situation in the "comments for editor" box. Not sure if that's the consensus approach...
1
0
2
@pedrohcgs
@dlmillimet
If I've whetted your appetite, "Specification Searches" is freely available online:
0
0
1
This merits more than 40 google scholar citations:
"A new analytic procedure, the highly touted Fractured Reconstructive Autoerotic Projection Package with Econometrization (FRAPPE) ... is used to detail the properties of the counterfactual world"
0
0
1
An R package to implement the methods from my recent JBES paper is now on CRAN thanks to the tireless efforts of
@EMallickHossain
#rstats
0
2
1
@Josh_Merfeld
@WeidnerEcon
This is the key question. Still working on it, but a couple thoughts (1/3)
1
0
1
@Apoorva__Lal
Yes you're totally right. So if we want the standard error for the ATE computed for these particular X-values in our sample, then we don't need to propagate uncertainty in X-bar. If we want them in the population of X-values we need to use GMM as you suggested. Thanks!
0
0
1
@jeffrlin
While aimed at mathematicians, the advice in Section 4.1 of Krantz's "Primer of Mathematical Writing" may be helpful:
0
0
1
Thanks to everyone who attended: it was a great session! If you missed it, it's not too late: we recorded the session and if you email me I can give you access along with adding you to the list for future meetings.
1
0
1
P.S. you too can make such a hexagon with minimal effort:
0
0
1
@rlmcelreath
Fun to learn that the intersection of music people I follow and stats people I follow is non-empty. Liszt prepared an edition of some of Field's nocturnes and wrote an introductory essay comparing w/ Chopin's
0
0
1
@mikedecr
Nice post! Recently been thinking about related research directions.
Also glad to see I'm not the only one who found the "Illusion of Learning" paper click-baity...
1
0
1
@tonyprinciotti
I enjoy a lot of Busoni's Bach transcriptions, but for the Chaconne I'm afraid it's got to
#Brahms
for the left-hand!
0
0
1
Seems too good to be true, but it looks like I've actually crossed all the teaching and admin items of my todo list for 2021. I think I'll just pretend 2022 doesn't exist for a couple of weeks...
0
0
1
@maxkasy
@UChicagoPress
I was lucky to get a peek at an early draft and can't wait for the final version!
0
0
1
@Josh_Merfeld
May I take this opportunity to advertise ? We're eager for feedback and hoping to announce some first steps soon.
@WeidnerEcon
1
0
1
@OxfordEconDept
Just finished teaching this course and had a blast! I tested out some new ideas for teaching selection-on-observables and IV and they seemed to work well. My materials are available here:
0
0
1
In "print" at long last:
0
0
1
Thanks to
@ClaireBolderson
and Caroline Finnigan for a great workshop on "Engaging With the Media." A prospect that used to make me feel uneasy now feels doable, even exciting!
1
0
1
Thanks to all our new signatories for your support! We're up to 96 signatures from A-Z. Who wants to push us over 100?
@WeidnerEcon
#sqare
#Econometrics
#econtwitter
0
0
1
@lijian44910521
@stat110
Excited to learn of the free online version: I've recommended it to my students.
1
0
1
Hot off the presses: a new working paper with co-authors in biostats at Cambridge & Imperial extending my past work on instrument selection and applying it to mendelian randomization:
Fun to learn about new (to me) applications of IV!
0
0
1
We estimate effectively zero spillovers on the untreated, but large negative spillovers on those willing to take the treatment. For this group, treatment has a protective effect, completely cancelling out the negative spillovers. This is consistent with selection on gains.(14/15)
1
0
1
It also makes sense given what we know about the treated: at baseline they are less educated, more likely to receive unemployment insurance, less likely to cohabit, etc. Thanks to
@rolandrathelot
for helping us access and understand this very cool dataset! (15/15)
0
0
1
When economists hear "non-compliance" they think of instrumental variables (IVs). The saturations and individual treatment offers in an RS experiment seem like ideal instruments. The question is how to use them. We begin by showing that the "obvious" idea doesn't work. (7/15)
1
0
1
Love this. Use it as an example in my course:
Just in time for the Mavericks to win the NBA Finals, I wrote about the surprising fact of probability that for decades made scientists wrongly think the "hot hand" was an illusion. For
@NautilusMag
1
4
10
0
0
1
@DrKorica
We're caught in a bad equilibrium in my field (econometrics) too. In case you're curious, is an attempt to change things that I'm working on with
@WeidnerEcon
0
0
1