Let’s learn about PyFixest and salary prediction models

Dear members of our community,

Hereby we announce our next meetup that takes place on August 29th and is hosted by trivago N.V.! In addition to a topic presented by trivago, this meetup will also incorporate economically highly relevant and data science based insights on salaries from Stepstone!

There will be two presentations. First, we will hear from Trivago’s Alexander Fischer who will present „Mostly Harmless Data Science: Linear Regression with PyFixest”.

Our next speakers from Stepstone Michael Matuschek and Tim Elfrink will present the topic of salary predictions at Stepstone.

Please find the detailed abstracts below.

Abstracts:

Speaker 1: Alexander Fischer, Data Scientist from trivago N.V

Title: Mostly Harmless Data Science Linear Regression (with high dimensional fixed effects) with PyFixest.

Abstract: In this talk, Alex will present his open source project PyFixest, a Python library that aims to bring the functionality, performance and ease of use of the outstanding “fixest” R package to the Python ecosystem. PyFixest supports fast estimation of a range of regression models with high dimensional fixed effects, including ordinary least squares (OLS) and instrumental variable (IV) estimation, along with various routines for conducting inference, including heteroscedasticity-robust (HC) and cluster robust standard errors and the wild (cluster bootstrap). Alex will discuss the package’s design principles and share a real-world use case from his work at trivago that demonstrates the practical applications of PyFixest).

Moreover, he will reflect on his experience using tools like ChatGPT and Copilot to develop code in his development process and differences in developing open source packages in R and Python.

github repo: https://github.com/s3alfisc/pyfixest

docs: https://s3alfisc.github.io/pyfixest/

Speakers 2: Michael Matuschek, Engineering & Data Science Lead and Tim Elfrink, ML Engineer from Stepstone GmbH

Title: The World of Salary: Introducing Salary Prediction Models at StepStone

Abstract: This presentation explores the salary landscape in the German job market, focusing on the challenges of data collection and approaches used to analyse it.

We will discuss the importance of getting just the right features and how to balance the amount of data used. We will also examine the pipeline from experimentation to production models and the importance of keeping track of metrics and how we can automatise the process.

Lastly, we will delve into the challenges of gender bias, data representation, and monotonicity, looking at how these factors impact our predictions, as well as prospects for future work.

By attending this meetup you consent to trivago's event privacy policy.