Fan Li @FanLiDuke Twitter profile

Pinned Tweet

Fan Li

5 months

Done another semester teaching causal inference🙂. Updated my course slides, added survival data, labs, corrected more typos this time. Close to 800 pages now. Always more to update next year.

12

148

658

Last Seen Profiles

@Nagoya3909

@TvBrada

@bDhM8033Ox7ASyG

@midnigh84569163

@KylieWe97185866

@Dominican_n1

@DexterCanal

@AAPOR

@pj_maine

@Hose503

@TheDylanMiles32

@stw_pdg

@infoGACuk

@bokeplokalmalam

@kyo_dir_en_gray

@Sommy_Dav

@nazu_kansoku

@bokeplokalmalam

@stwmaniax

@YumeKemuri__

@anesthesiadocmd

@sobredxsis

@deportesDDS

@Goingparadisee

@Elizeud72248412

@Hoshino_Fox

@TheRealRynnstar

@MrsParkOTHS

@Queenofpride2

@stwmaniax

@ungeorgia

@bokeplokalmalam

@Billtrust

@stw_pdg

@KaiTheFlower1

Fan Li

@FanLiDuke

3 months

Twitter academia: 1. I am happy to announce xx (whatever trivial) 2. I am thrilled/excited that xx (papers, grants, promotion) 3. I am honored that xx ("awards" in all senses) Adding to that list now: "How I publish xx papers in x years" What next? How I become god?

52

135

2K

Fan Li

@FanLiDuke

2 months

Hands-down the best textbook for causal inference.

Peng Ding

@pengding00

2 months

excited to see the physical copy of the book; nervous about the potential errors. Comments are welcome.

24

117

772

1

16

233

Fan Li

@FanLiDuke

5 months

Lack of general code is one main barrier to Bayesian causal inference. Peng Ding gave some simple Stan code and intro slides on Bayesian causal inference based on our review paper. A long version is here:

Peng Ding

@pengding00

5 months

Just gave a guest lecture on Bayesian Causal Inference at Williams College, with slides and R code at which is an introduction to our review paper @fabri_mealli @FanLiDuke (I never taught any Bayesian Statistics at Berkeley.)

0

58

207

1

26

127

Fan Li

@FanLiDuke

6 months

Happy to see my paper with @pengding00 and Anqi Zhao @DukeU "Covariate adjustment in randomized experiments with missing outcomes and covariates" is out This 8-pages paper gives a simple and clean solution to a prevalent practical problem.

Covariate adjustment in randomized experiments with missing outcomes and covariates

Summary. Covariate adjustment can improve precision in analysing randomized experiments. With fully observed data, regression adjustment and propensity sco

academic.oup.com

1

31

94

Fan Li

@FanLiDuke

5 months

Another great book draft (on linear models) from Peng Ding @pengding00 . Clean, clear, and concise. Very up-to-date.

Kirk Borne

@KirkDBorne

5 months

[Download 400-page PDF eBook] Linear Models in #MachineLearning and Statistical Learning: ————— #Mathematics #DataScience #LinearAlgebra #Statistics

1

74

287

0

10

67

Fan Li

@FanLiDuke

3 months

@MarvinSchmittML and of course, wanting the whole world to know "I am humbled"

2

58

Fan Li

@FanLiDuke

5 months

@francoisfleuret No model is causal, assumptions make it "causal." Blame the abuse of the name "causal model" and the lack of assumption checking instead.

2

6

37

Fan Li

@FanLiDuke

5 months

Excellent point. Serving as the editor for Social Science, Biostatistics, Policy at Annals of Applied Statistics, I have spent countless hours reviewing papers, burned social capitals, offending many authors (rejecting their papers). My stipend? 0. Is the system sustainable?

Jason Fletcher

@jasonmfletcher

5 months

I am honored to be offered the job of Editor-in-Chief of Health Economics @HECJournalTweet , replacing the excellent Sally Stearns ( @GillingsGlobal ). I plan to reject this offer. Why?

6

114

389

0

1

29

Fan Li

@FanLiDuke

3 months

@economeager "How I become god" is a even better click bait than "(non econ) how I write xx papers in x years” or "(econ) how I write one paper in less than 10 years."

2

0

18

Fan Li

@FanLiDuke

4 months

@5_utr Reality is, even the best designed RCT, at the time of completion, looks more like observational studies, because of, e.g. unintentional and uncontrolled intercurrent events. The blanket hostility to obs studies is unwarranted and a dis-service

2

1

14

Fan Li

@FanLiDuke

5 months

Good stat joke. Once in a party, the late Susie Bayarri (one of the great Bayesians) and a few of us were discussing who is the statistician divorced the most times. Susie said: "He must be a frequentist!"

2

1

14

Fan Li

@FanLiDuke

5 months

Saw many cicada shells everywhere lately. Turned out two different broods of cicadas (one on a 13 yr and other on a 17 yr cycle -- two prime numbers) emerge at the same time from underground this year, first time since 1803, next time? 2024+17X13=2245.

0

12

Fan Li

@FanLiDuke

5 months

Brilliant strike back. I was compelled to google who Genc is.

Ian P. McCarthy

@Toffeemen68

5 months

The peer review process.

63

386

4K

0

1

13

Fan Li

@FanLiDuke

5 months

Great new paper on the controversial issue of causal interpretation of hazard ratio by the Yale Fan @FanLi90

Brennan Kahan

@Brennan_Kahan

5 months

Really interesting new paper by @FanLi90 and Michael Fay on the causal interpretation of the hazard ratio in RCTs. Also lots of great stuff on estimands more generally.

2

18

50

0

12

Fan Li

@FanLiDuke

3 months

@ssprickschuster Except that often "I am humbled/honored" is a simply thin-veiled bragging of trivial things or openly flattering relevant people (e.g. potential reviewers or letter writers)

1

0

12

Fan Li

@FanLiDuke

6 months

R21's are hard!

0

1

10

Fan Li

@FanLiDuke

4 months

@BhramarBioStat Second this. Learning some history of statistics (started from survey sampling) also helps.

1

0

8

Fan Li

@FanLiDuke

5 months

@f2harrell The key factor is whether the treatment is the same within the same cluster; if so, observational studies also pay a as big if not bigger price as CRT

0

6

Fan Li

@FanLiDuke

6 months

An interesting talk tomorrow "A Measure-Theoretic Axiomatisation of Causality" at Online Causal Inference Seminar Axiomatic foundation of statistics is still arguably murky with exception of Bayesian. Question is do we really need causal axiomatisation?

1

0

6

Fan Li

@FanLiDuke

6 months

@ZhenkeWu I got assigned 14 papers on the first month as AOAS area editor. So way to go for you 🤣

0

5

Fan Li

@FanLiDuke

4 months

@noah_greifer @stephensenn A caveat: the asymptotic equivalence holds for IPW, but for OW only with r=0.5 (r: trt proportion). We added a correction recently.

0

4

Fan Li

@FanLiDuke

5 months

@pedrohcgs @noah_greifer Copying @noah_greifer , of course our own R package PSweight 🤣

0

4

Fan Li

@FanLiDuke

4 months

@f2harrell @yudapearl @soboleffspaces @stephensenn @PWGTennant In fact, most CI practitioners estimate a third estimand, a mix of PATE and SATE, because of the conditioning on the X in the sample. My Bayesian causal review paper carefully discussed this. In practice, diff between P-, S-, M- ATE is small, not warrant much concern

2

0

3

Fan Li

@FanLiDuke

4 months

@f2harrell @PWGTennant great idea, will adopt later.

0

2

Fan Li

@FanLiDuke

5 months

@Brennan_Kahan @FanLi90 This is great and timely work

0

2

Fan Li

@FanLiDuke

2 months

@sasilu6 What's wrong with lecture notes? This gives the most concise, update-to-date and complete survey of the field, with examples, HWs, codes. Any other book on market offers these?

1

0

2

Fan Li

@FanLiDuke

5 months

In a hotel today, I was wondering how much and how people leave tips for hotel cleaning. Most people, like me, don't carry small bills. Does this hurt cleaning staff's income?

1

0

2

Fan Li

@FanLiDuke

5 months

@f2harrell @BrendenDufault Clusters come in nature in obs studies: hospitals, insurance plans, physicians. The question is clustering in what sense? I think the answer lies in the correlation of the outcomes between units in a cluster, breaking iid assumption.

0

1

Fan Li

@FanLiDuke

6 months

@DG32163630 My colleague Jim Berger had nice discussions about this in his class on history of (Bayesian) statistics. Not sure if it is written up. For some references, see the old debates in the foundation of stat, e.g., arising from Birnbaum's landmark paper on the Likelihood Principle

1

0

1

Fan Li

@FanLiDuke

5 months

@pedrohcgs @noah_greifer PSweight doesn't have HT option. Because it is written for IPW, ATT, OW, among others, and OW requires normalization, we decided to go straight Hajek. Maybe it is time to add HT

1

0

1

Fan Li

@FanLiDuke

4 months

@f2harrell @yudapearl @soboleffspaces @stephensenn @PWGTennant CATE is a different animal, fundamentally you lack data to estimate infinite number of CATEs, of course the uncertainty is bigger. I don't think you can get a good sample CATE in general.

0

Fan Li

@FanLiDuke

5 months

@inferencelab That's great! Marked

0

1

Fan Li

@FanLiDuke

4 months

@lhan320 So sorry for your loss

0

1

Fan Li

@FanLiDuke

4 months

@gv_lazcano @5_utr I am not saying one should go to obs studies. I am saying, to analyze your holy grail RCTs properly, many causal inference techniques designed for observational studies are necessary. Again, blanket hostility is counter-productive.

2

0

1

Fan Li

@FanLiDuke

4 months

@yudapearl @f2harrell @soboleffspaces @stephensenn @PWGTennant There is a subtle technical difference between the three versions of ATE, but I don't believe the technical purity of PATE and SATE makes them more superior. In practice, the numerical difference is little - that matters the most.

0

1