Causal Inference

Causal inference allows us to connect the dots in our data and speak to the cause-and-effect relationship that one thing may have with another. In ideal conditions, a thorough experiment design will be repeatable to a high degree of consistency. In practice, however, many fields that apply data science require some form of cause-and-effect recognition solely from observational data. To do this, it is imperative to understand causality, how to define a causal research question, and what causal models can and cannot tell us from observational data.

The first two sections of this course focus on the why and what of research design, covering key motivations and concepts. The third and fourth sections provide you with an introduction to the technical side. You will learn how common causal models work, what they are meant to achieve, how to apply them, and perhaps most importantly, their limitations.

Requirements

This course is designed for those who have a degree in something other than Computer Science/Statistics and are looking to enhance their data science skills for their career.

Learning Outcomes

  • The ability to understand the concepts of research design and causality
  • The ability to formulate a suitable research question
  • How to structure a robust experimental study
  • The ability to understand the benefits and drawbacks of causal inference models
  • How to implement common models and techniques
  • The ability to understand and think critically about causal inference with observational data

Delivery Format and Schedule

Online for 7 hours/week for 3 weeks (21 hours in total).

2023 Dates

  • Monday 3 April, 6pm-8pm: Research basics I (designing research; research questions)
  • Thursday 6 April, 6pm-8pm: Research basics II (variables and relationships; identification)
  • Saturday 8 April, 9am-noon: Causality I (causal diagrams; causal paths)
  • Monday 10 April, 6pm-8pm: Causality II (treatment effects; rules of thumb)
  • Thursday 13 April, 6pm-8pm: Methods I (regression; matching; fixed effects)
  • Saturday 15 April, 9am-noon: Methods II (differences; instrumental variables; discontinuity)
  • Monday 17 April, 6pm-8pm: Reproducibility; ethics; inequity
  • Thursday 20 April, 6pm-8pm: Professional skills: Industry case study
  • Saturday 22 April, 9am-noon: Causal Inference: Review and Practice