Ofir Reich (@OfirReich) header
Ofir Reich (@OfirReich) profile picture

Ofir Reich

@OfirReich

Data Scientist. Africa enthusiast.

Joined March 2017
  • Tweets152
  • Following48
  • Followers195
33Media
Photo by Ofir Reich (@OfirReich)Photo by Ofir Reich (@OfirReich)Photo by Ofir Reich (@OfirReich)Photo by Ofir Reich (@OfirReich)Photo by Ofir Reich (@OfirReich)Photo by Ofir Reich (@OfirReich)Photo by Ofir Reich (@OfirReich)Photo by Ofir Reich (@OfirReich)Photo by Ofir Reich (@OfirReich)Photo by Ofir Reich (@OfirReich)
World Cup player origins. See all at: ofirrei.ch/world-cup-originsโ€ฆ ๐Ÿ‡จ๐Ÿ‡ญ SWITZERLAND ๐Ÿ‘ถ๐Ÿ‡จ๐Ÿ‡ญ๐Ÿ‡จ๐Ÿ‡ญ๐Ÿ‡จ๐Ÿ‡ญ๐Ÿ‡จ๐Ÿ‡ญ๐Ÿ‡จ๐Ÿ‡ญ๐Ÿ‡จ๐Ÿ‡ญ๐Ÿ‡จ๐Ÿ‡ญ๐Ÿ‡จ๐Ÿ‡ญ๐Ÿ‡จ๐Ÿ‡ญ๐Ÿ‡จ๐Ÿ‡ฒ๐Ÿ‡จ๐Ÿ‡ญ ๐Ÿ‡ซ๐Ÿ‡ท FRANCE ๐Ÿ‘ถ๐Ÿ‡ฌ๐Ÿ‡ซ๐Ÿ‡ซ๐Ÿ‡ท๐Ÿ‡ซ๐Ÿ‡ท๐Ÿ‡ซ๐Ÿ‡ท๐Ÿ‡ซ๐Ÿ‡ท๐Ÿ‡ซ๐Ÿ‡ท๐Ÿ‡ซ๐Ÿ‡ท๐Ÿด๓ ง๓ ข๓ ฅ๓ ฎ๓ ง๓ ฟ๐Ÿ‡ซ๐Ÿ‡ท๐Ÿ‡ซ๐Ÿ‡ท๐Ÿ‡ซ๐Ÿ‡ท ๐Ÿ‡ฒ๐Ÿ‡ฆ MOROCCO ๐Ÿ‘ถ๐Ÿ‡จ๐Ÿ‡ฆ๐Ÿ‡ช๐Ÿ‡ธ๐Ÿ‡ช๐Ÿ‡ธ๐Ÿ‡ซ๐Ÿ‡ท๐Ÿ‡ณ๐Ÿ‡ฑ๐Ÿ‡ซ๐Ÿ‡ท๐Ÿ‡ซ๐Ÿ‡ท๐Ÿ‡ช๐Ÿ‡ธ๐Ÿ‡ง๐Ÿ‡ช๐Ÿ‡ฒ๐Ÿ‡ฆ๐Ÿ‡ช๐Ÿ‡ธ ๐Ÿ‡ฉ๐Ÿ‡ช GERMANY ๐Ÿ‘ถ๐Ÿ‡ฉ๐Ÿ‡ช๐Ÿ‡ฉ๐Ÿ‡ช๐Ÿ‡ฉ๐Ÿ‡ช๐Ÿ‡ฉ๐Ÿ‡ช๐Ÿ‡ฉ๐Ÿ‡ช๐Ÿ‡ฉ๐Ÿ‡ช๐Ÿ‡ฉ๐Ÿ‡ช๐Ÿ‡ฉ๐Ÿ‡ช๐Ÿ‡ฉ๐Ÿ‡ช๐Ÿ‡ฉ๐Ÿ‡ช๐Ÿ‡ฉ๐Ÿ‡ช ๐Ÿ‡ธ๐Ÿ‡ณ SENEGAL ๐Ÿ‘ถ๐Ÿ‡ซ๐Ÿ‡ท๐Ÿ‡ธ๐Ÿ‡ณ๐Ÿ‡ซ๐Ÿ‡ท๐Ÿ‡ซ๐Ÿ‡ท๐Ÿ‡ธ๐Ÿ‡ณ๐Ÿ‡ธ๐Ÿ‡ณ๐Ÿ‡ซ๐Ÿ‡ท๐Ÿ‡ธ๐Ÿ‡ณ๐Ÿ‡ธ๐Ÿ‡ณ๐Ÿ‡ธ๐Ÿ‡ณ๐Ÿ‡ฌ๐Ÿ‡ฒ
1
1
122
more replies
๐Ÿ‡ฟ๐Ÿ‡ฆ SOUTH AFRICA ๐Ÿ‘ถ๐Ÿ‡ฟ๐Ÿ‡ฆ๐Ÿ‡ฟ๐Ÿ‡ฆ๐Ÿ‡ฟ๐Ÿ‡ฆ๐Ÿ‡ฟ๐Ÿ‡ฆ๐Ÿ‡ฟ๐Ÿ‡ฆ๐Ÿ‡ฟ๐Ÿ‡ฆ๐Ÿ‡ฟ๐Ÿ‡ฆ๐Ÿ‡ฟ๐Ÿ‡ฆ๐Ÿ‡ฟ๐Ÿ‡ฆ๐Ÿ‡ฟ๐Ÿ‡ฆ๐Ÿ‡ฟ๐Ÿ‡ฆ ๐Ÿ‡ฐ๐Ÿ‡ท SOUTH KOREA ๐Ÿ‘ถ๐Ÿ‡ฐ๐Ÿ‡ท๐Ÿ‡ฐ๐Ÿ‡ท๐Ÿ‡ฐ๐Ÿ‡ท๐Ÿ‡ฐ๐Ÿ‡ท๐Ÿ‡ฐ๐Ÿ‡ท๐Ÿ‡ฐ๐Ÿ‡ท๐Ÿ‡ฐ๐Ÿ‡ท๐Ÿ‡ฐ๐Ÿ‡ท๐Ÿ‡ฐ๐Ÿ‡ท๐Ÿ‡ฐ๐Ÿ‡ท๐Ÿ‡ฐ๐Ÿ‡ท ๐Ÿ‡ช๐Ÿ‡ธ SPAIN ๐Ÿ‘ถ๐Ÿ‡ช๐Ÿ‡ธ๐Ÿ‡ช๐Ÿ‡ธ๐Ÿ‡ซ๐Ÿ‡ท๐Ÿ‡ซ๐Ÿ‡ท๐Ÿ‡ช๐Ÿ‡ธ๐Ÿ‡ช๐Ÿ‡ธ๐Ÿ‡ช๐Ÿ‡ธ๐Ÿ‡ช๐Ÿ‡ธ๐Ÿ‡ช๐Ÿ‡ธ๐Ÿ‡ช๐Ÿ‡ธ๐Ÿ‡ช๐Ÿ‡ธ ๐Ÿ‡ธ๐Ÿ‡ช SWEDEN ๐Ÿ‘ถ๐Ÿ‡ธ๐Ÿ‡ช๐Ÿ‡ธ๐Ÿ‡ช๐Ÿ‡ธ๐Ÿ‡ช๐Ÿ‡ธ๐Ÿ‡ช๐Ÿ‡ธ๐Ÿ‡ช๐Ÿ‡ธ๐Ÿ‡ช๐Ÿ‡ธ๐Ÿ‡ช๐Ÿ‡ธ๐Ÿ‡ช๐Ÿ‡ธ๐Ÿ‡ช๐Ÿ‡ธ๐Ÿ‡ช๐Ÿ‡ธ๐Ÿ‡ช ๐Ÿ‡น๐Ÿ‡ณ TUNISIA ๐Ÿ‘ถ๐Ÿ‡น๐Ÿ‡ณ๐Ÿ‡ซ๐Ÿ‡ท๐Ÿ‡ซ๐Ÿ‡ท๐Ÿ‡ฉ๐Ÿ‡ช๐Ÿ‡ซ๐Ÿ‡ท๐Ÿ‡ซ๐Ÿ‡ท๐Ÿ‡ซ๐Ÿ‡ท๐Ÿ‡ฉ๐Ÿ‡ช๐Ÿ‡ซ๐Ÿ‡ท๐Ÿ‡ณ๐Ÿ‡ด๐Ÿ‡น๐Ÿ‡ณ ๐Ÿ‡น๐Ÿ‡ท TรœRKIYE ๐Ÿ‘ถ๐Ÿ‡น๐Ÿ‡ท๐Ÿ‡น๐Ÿ‡ท๐Ÿ‡น๐Ÿ‡ท๐Ÿ‡น๐Ÿ‡ท๐Ÿ‡ณ๐Ÿ‡ฑ๐Ÿ‡ฉ๐Ÿ‡ช๐Ÿ‡ณ๐Ÿ‡ฑ๐Ÿ‡น๐Ÿ‡ท๐Ÿ‡น๐Ÿ‡ท๐Ÿ‡ฉ๐Ÿ‡ช๐Ÿ‡น๐Ÿ‡ท ๐Ÿ‡บ๐Ÿ‡พ URUGUAY ๐Ÿ‘ถ๐Ÿ‡ฆ๐Ÿ‡ท๐Ÿ‡บ๐Ÿ‡พ๐Ÿ‡บ๐Ÿ‡พ๐Ÿ‡บ๐Ÿ‡พ๐Ÿ‡บ๐Ÿ‡พ๐Ÿ‡บ๐Ÿ‡พ๐Ÿ‡บ๐Ÿ‡พ๐Ÿ‡บ๐Ÿ‡พ๐Ÿ‡บ๐Ÿ‡พ๐Ÿ‡บ๐Ÿ‡พ๐Ÿ‡บ๐Ÿ‡พ
1
1
36
๐Ÿ‡บ๐Ÿ‡ธ UNITED STATES ๐Ÿ‘ถ๐Ÿ‡บ๐Ÿ‡ธ๐Ÿ‡ณ๐Ÿ‡ฑ๐Ÿ‡บ๐Ÿ‡ธ๐Ÿ‡บ๐Ÿ‡ธ๐Ÿด๓ ง๓ ข๓ ฅ๓ ฎ๓ ง๓ ฟ๐Ÿ‡บ๐Ÿ‡ธ๐Ÿ‡บ๐Ÿ‡ธ๐Ÿ‡บ๐Ÿ‡ธ๐Ÿ‡ฉ๐Ÿ‡ช๐Ÿ‡บ๐Ÿ‡ธ๐Ÿ‡บ๐Ÿ‡ธ ๐Ÿ‡บ๐Ÿ‡ฟ UZBEKISTAN ๐Ÿ‘ถ๐Ÿ‡ฐ๐Ÿ‡ฟ๐Ÿ‡บ๐Ÿ‡ฟ๐Ÿ‡บ๐Ÿ‡ฟ๐Ÿ‡บ๐Ÿ‡ฟ๐Ÿ‡บ๐Ÿ‡ฟ๐Ÿ‡บ๐Ÿ‡ฟ๐Ÿ‡บ๐Ÿ‡ฟ๐Ÿ‡บ๐Ÿ‡ฟ๐Ÿ‡บ๐Ÿ‡ฟ๐Ÿ‡บ๐Ÿ‡ฟ๐Ÿ‡บ๐Ÿ‡ฟ
1
16
Have you noticed the word Fable sounds a lot like Sable? #IfAnyoneBuildsIt
2
1
1
8
1,182
New paper: "Prediction decomposition for causal analysis" Using ML predictions as RCT outcomes can give a treatment effect of ZERO โ€” even with great test-set Rยฒ. I show why and propose a fix. arxiv.org/abs/2604.11168 ๐Ÿงต...
Image from Ofir Reich (@OfirReich)
1
1
39
more replies
Two uses: 1) Model selection: pick highest ฮท_ฮต, not highest Rยฒ. No strong assumptions. 2) Bias correction: divide effect by ฮทฬ‚_ฮต. Needs ฮท_T โ‰ˆ ฮท_ฮต.
Image from Ofir Reich (@OfirReich)
1
16
Paper + open code: arxiv.org/abs/2604.11168 github.com/ofir-reich/predicโ€ฆ
Prediction decomposition for causal analysis

Prediction decomposition for causal analysis

There is rising interest in using Machine Learning (ML) model predictions as outcomes in causal analysis. However, these methods have faced challenges in finding the true treatment effects. It is...

arxiv.org
7
exactly reddit.com/r/xkcd/comments/1โ€ฆ
Image from Ofir Reich (@OfirReich)
47
โ€œIs Model Accuracy Enough? A Field Evaluation Of A Machine Learning Model To Catch Bogus Firms.โ€ New paper: #MachineLearning model accuracy need not translate to causal impact. W/ Taha Barwahwala, @aprajitmahajan, @shekhar_m. ๐Ÿ™ @JPAL_SA ๐Ÿ™ @ICTDTax. dl.acm.org/doi/abs/10.1145/3โ€ฆ ๐Ÿงต
Image from Ofir Reich (@OfirReich)
1
1
4
246
more replies
How come no treatment effect? A) Overfitting to proxy-labels. Model predicts firms already on a trajectory to be canceled, e.g. havenโ€™t filed returns for months. B) Lacking department enforcement. Firms non-existent in inspections, likely not canceled more than 1 year later.
Image from Ofir Reich (@OfirReich)
1
1
49
Implications. Model accuracy is not enough. Field evaluation is necessary to determine real-world impact on existing system. Other ICT solutions to lubricate enforcement action and collaboration within the department may be beneficial. Paper again: dl.acm.org/doi/abs/10.1145/3โ€ฆ
Is Model Accuracy Enough? A Field Evaluation of a Machine Learning Model to Catch Bogus Firms | ACM...

Is Model Accuracy Enough? A Field Evaluation of a Machine Learning Model to Catch Bogus Firms | ACM...

We investigate the use of a machine learning (ML) algorithm to identify fraudulent non-existent firms that are used for tax evasion. Using a rich dataset of tax returns in an Indian state over...

dl.acm.org
1
31
Load more