Getting Started with Quasi-Experiments in Data Science

Description:

Getting Started with Quasi-Experiments in Data Science

Each year businesses spend millions on consumer discounts, employee training, loyalty programs, and other similar interventions. They turn to data scientists to test if these programs are effective, learn how to improve them, and decide when to scrap them. Problematically, randomized controlled trials (RCTs), which are considered the “gold standard” for analyzing interventions, are usually implausible in practice.

This is where quasi-experimental techniques, like propensity score matching and interrupted time-series design, become valuable tools in the data scientists’ arsenal. Since quasi-experiments do not require randomization (or even control groups), they can establish causal relationships when RCTs aren’t possible, allowing data scientists to rise to the challenge these interventions pose.

To help attendees start on the path to using these techniques, this presentation aims to do four things:

1. Provide an easy-to-follow introduction to quasi-experiments and a discussion of common criticisms and rebuttals.
2. Propose a simple framework for identifying causal problems and pairing them with suitable quasi-experimental designs.
3. Give an overview of the key techniques in the quasi-experimental design toolbox and what is needed to apply them successfully.
4. Carry out an applied example with focus on interpreting and communicating the findings to citizen stakeholders.

The goal of this presentation is to provide beginners a first step into the world of quasi-experimental design, while offering more experienced participants a widened view of the area and a simple framework they can take back to less experienced colleagues.

Speaker:
Anthony Melson is Senior Manager of Data Science at Maritz and one of the co-organizers of St Louis Machine Learning and Data Science. As a data scientist, he has experience applying causal analysis, experimental design, and machine learning at some of the largest companies in the world. As a manager, he leads a team of 9 data scientists and focuses much of his time on training and mentoring junior data scientists.

Event Details (In-person/Online):
This is a hybrid event. People can either attend to the in-person location or online through zoom.

In-person:
The in-person event will be held at the Slalom office on the 8th floor, suite 850.
We will have people in the lobby to ensure you can enter the building and the office. We encourage you to come to the office.

Parking details:
On-street parking is free after 5 PM on weekdays.
Food and refreshment will be provided for the in-person event!

Online:
The meeting URL should bypass the prompts to input the meeting ID and password. However, full meeting details are provided below.

Meeting ID: 996 3956 4420

Meeting Password: stlmlds

Join Zoom Meeting with this URL (bypasses ID and password):
https://slalom.zoom.us/j/99639564420?pwd=ckdNaENSWE4wNEZNMU4ycCtjeTA1QT09

Networking - 6:30 PM to 7:00 pm
Presentation - 7:00 PM to 7:45 PM
Q&A - 7:45PM to 8:00 PM
Networking - 8:00 PM to 8:30PM