Ami Wulf

Data Science @ Etsy, Stats PhD from UCLA, causal inference, publishing under David Ami Wulf. These opinions have no unblocked backdoor paths to anyone but me.

[Ami Wulf @DAmiWulf]

Back in June, I got my PhD. Below are some words of thanks and reflection, and links to some other 🧵

[Ami Wulf @DAmiWulf]

Some insightful connections from @stats_samir related to work he was doing at Etsy this summer. Clearly we were lucky to have him as our intern!

[Samir Khan @stats_samir]

A thread about CUPAC/CUPED - the tl;dr is that "CUPAC" is nearly equivalent to G-computation/Oaxaca-Blinder/imputation methods with a somewhat unusual choice of imputation model

https://t.co/tKF3JlfUdo

1/10

[Ami Wulf @DAmiWulf]

Wrote a blog post for @Etsy about a project I work on called interleaving — like if a paired test version of A/B experiments met a shuffled deck of cards… plus there are some pictures of weird clocks in there too #techblog #statstwitter #interleaving #ml

[Etsy Engineering @codeascraft]

Have lots of product and #MachineLearning ideas but not enough online traffic to test them all? Sr. Data Scientist @DAmiWulf shares how #interleaving lets teams at @Etsy run certain experiments using < 10% of the traffic needed for A/B tests #DataScience https://t.co/h9sVKG3Bxp

[Ami Wulf @DAmiWulf]

Had the opportunity to tag along with some great researchers on this timely paper. Thanks to @RebeccaKMiller1 for the leadership!
"Our results indicate a potentially fundamental policy shift in response to recent wildfires in California."

[Dr. Rebecca Miller @RebeccaKMiller1]

📰📢 How has #wildfire policy in California changed over time? Read our new paper out in @IOPenvironment “Trends in wildfire legislation in California, 2001-2020” to find out! 🔥 more details below: https://t.co/wA9iKeP29C

[Ami Wulf @DAmiWulf]

@Teich50 @marketsensei This table might be helpful https://t.co/eF6omoT92x

Beta = False Negative Rate = 1 - True Positive Rate = 1 - power
while
False Discovery Rate = 1 - precision = 1 - Positive Predictive Value

There is no equation relating power and FDR that holds across applications

[Ami Wulf @DAmiWulf]

Things that haven’t successfully got me to learn the Greek alphabet:
-frat row in undergrad
-math & stats coursework
-LaTeX
-wanting to not embarrass myself asking questions in seminars

Next challenger:
COVID variants other than alpha, epsilon, mu, pi, tau, sigma, or chi

[Meg Tirrell @megtirrell]

It’s Omicron

[Ami Wulf @DAmiWulf]

It’s Bayes all the way down

[xkcd @xkcd_rss]

Bayes' Theorem https://t.co/HGgD7DbAOo

[Hilary Parker @hspter]

The perfect tweet

[Ami Wulf @DAmiWulf]

(And by “old” Bay Area, I mean “slightly less new” Bay Area)

[Ami Wulf @DAmiWulf]

Whether you first confused the transit system BART with BART the ensemble-of-trees model or BART the NLP autoencoder is a great measure of whether you’re old Bay Area or new Bay Area. Auto-disqualification if you say “The BART” though

[BART @SFBART]

If you haven't mixed us, a public transit system, up with a nonparametric Bayesian regression approach which uses dimensionally adaptive random basis elements, are you even from the Bay?

[Ami Wulf @DAmiWulf]

Forget Red (Taylor’s version), it’s all about these classics

[Ami Wulf @DAmiWulf]

Var[X/Y] (Taylor’s version)
Gradient descent (Taylor’s version)
Linearization (Taylor’s version)

[Melody Huang @melodyyhuang]

Excited to share that a paper I've been working on with Naoki Egami, Erin Hartman, and Luke Miratrix (@LMiratrix) is on ArXiv now!
https://t.co/AdgqC4udws
We develop a new method to improve the generalization of experimental results (1/2)

[Valentin Amrhein @vamrhein]

One of the worst confusions about the P-value I have ever seen printed.
https://t.co/2Iv1uTWlq3

[Ami Wulf @DAmiWulf]

I highly recommend checking out this talk tomorrow by my UCLA stats officemate–– I mean by esteemed University of Washington Assistant Professor Carlos Cinelli

[Análise Real @analisereal]

Hi all, tomorrow (Tuesday) I will give a talk about sensitivity analysis in the Online Causal Inference Seminar. Hope to see you there! https://t.co/3CVjTDCJm1

[Pablo Geraldo B. @pablogerbas]

The amazing @bernardjkoch just posted on Arxiv "Deep Learning of Potential Outcomes". If you're either familiar with causality but not with deep learning, or the other way around, this is a great place to start!

https://t.co/CQBRPQ9yqQ

[Ami Wulf @DAmiWulf]

@nickchk @VincentAB As a causal inference researcher working for Etsy, this makes me very happy!

[Ami Wulf @DAmiWulf]

Really enjoyable day to be a disciple of these giants of causal inference

[The Nobel Prize @NobelPrize]

BREAKING NEWS:
The 2021 Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel has been awarded with one half to David Card and the other half jointly to Joshua D. Angrist and Guido W. Imbens.

#NobelPrize

[Jonathan Mummolo @jonmummolo]

Our new working paper: tell the computer what you know (and don't know!) about a causal question w/ discrete data → automatically get most precise possible answer (bounds, or a point estimate). Joint w/ @guilhermejd1 @nsfinkelstein @dean_c_knox Shpitser.🧵https://t.co/9FY4HKCp32

[Ami Wulf @DAmiWulf]

@nickchk Perhaps collider bias? Your effort/care E and the paper idea’s novelty/quality N both have arrows to both whether you submit S and to reviewer opinion O. Conditioning on S correlates E and N, which causes the inverted-U if you submit on E or N and the reviewers want both E & N

[Ami Wulf @DAmiWulf]

@graemedblair Thanks Graeme!

[Ami Wulf @DAmiWulf]

Back in June, I got my PhD. Below are some words of thanks and reflection, and links to some other 🧵

[Ami Wulf @DAmiWulf]

How it started How it ended

[Ami Wulf @DAmiWulf]

🧵 with advice — If you have followers starting off/considering grad school for stats or something similar, especially those looking to industry after graduation, I’d appreciate you sharing with them

[Ami Wulf @DAmiWulf]

I recently finished my PhD in statistics and started a dream data scientist role. If you’re starting a stats/stats-adjacent grad program (MS or PhD) I have some advice, particularly if you’re considering going into industry 🧵