29-30 Sep 21
September 29, 2021 - September 30, 2021
9:00 am - 5:00 pm
About the course
When using data analyses to steer decision-making, we want our actions to have the desired causal effect. And we all know that correlation does not imply causation. However, most techniques are limited to an associative, rather than a causal relationship. This training will broaden your analytical toolkit with different causal inference techniques. You will be taken on a full learning journey, beginning on the first day with the importance and value of causal relationships in business, answering the puzzling ‘what does this mean’ questions. We revise traditional statistics in order to make statistical claims about causality, using methods such as Randomized Trials and Instrumental Variables. On the second day, we cover conceptual causal models and get hands-on experience performing estimation with do calculus, one of today’s most exciting and promising ideas in this field. After this training, you will be able to assess causal effects and become more critical towards other analysis conclusions and claims about causality, which will lead to a greater impact with your analyses and business actions.
Why this is for you
In business we often want to know the effect of certain actions so we can take the best approach when making decisions. However, traditional statistics fail to give causal answers: drivers in regression are not causal, machine learning is just curve-fitting. Therefore, this causal inference course is crucial. It will allow you to translate real-world problems into a structural form and, by creating a causal model, estimate the effect of business interventions.
This training is perfect for Data Scientists or Data Engineers looking to formulate causal inference skills and put into practice theoretical knowledge for useful business applications. To make the step up to causality a strong statistical and mathematical background is required. Participants must also have previously conquered badges: 3201 Machine Learning Process, 3202 Classification Using Tree Models, and 3203 Regression Models.
What you’ll learn
This training will work with the concept of causal inference through the stages of revisioning, conceptual models, and estimation. Specifically, these three stages will include:
- Creating a directed acyclic graph
- Curve fitting
- Distilling testable implications from a causal model
- Using instrumental variables
- Applying randomized trials and treatment effects
- Constructing a Directed Acyclic Graph (DAG)
- Using a Causal Lift package in Python
- Estimating intervention effects using do calculus
- Importance: Able to explain the need and value of identifying causal relationships in business.
- Traditional statistics: Understanding the limitations and shortcomings of traditional statistical and machine learning methods when aiming for causal claims.
- Revisioning traditional statistics for causality: Being able to recognize when instrumental variables should be used and how to apply them
- Conceptual causal models: Being able to construct conceptual causal models
- Estimation in causal models: Being able to use a causal framework like Structural Causal Model and to estimate intervention effects using do-calculus
Theory and practical use
All trainings in the GAIn portfolio combine high-quality standardized training material with theory sessions from experts and hands-on experience where you directly apply the material to real-life cases. Each training is developed by top of the field practitioners which means they are full of industry examples along with practical challenges and know-how, fueling the interactive discussions during training. We believe this multi-level approach creates the ideal learning environment for participants to thrive.