The wise see trouble coming and prepare, but the simple keep going and are harmed. - Proverbs 27:12

Science and mathematics progress by degrees, sometimes by massive leaps. The progression is not something that you can know looking forward, it must be understood looking backward. Nor is progress necessarily linear. This holds true especially in our time of massive computation and parallelization. As progress conforms more to super linear laws, our ability to track it decreases. Yet we still desire to understand how far we’ve come.

For the last three to four months, I have been building a new company. I got introduced to a few entrepreneurs local to my area who had an idea for a business in the esports space. They needed someone technical to come on board for the vision and execute the product. That became me.

Here in Causal Inference Part 2, we go through a code implementation of causal inference applied to a healthcare problem. The problem was originally explored with some older, more traditional statistical techniques. Consider this the logical follow up to part 1 where we implement the mathematics described there.

Like waking up the day after a 20 mile hike, he opens his eyes. Only he didn’t go on a hike yesterday, too much work. Many questions will come to his head, once it has seen more sunlight. He knows he stares at the artificial sun too long before bed. It is a habit that he pretends he wants to shake. Tonight, he’ll do better. For now, he needs an IV drip of caffeine. The Monster Energy drinks of his college days are no longer an option, his digestive tract has “matured”. He feels the rough carpet under his feet as he transitions from prone to standing. The walk downstairs isn’t bad. He’s awake, just not completely aware yet. The sound of the coffee brewing triggers a warm sensation in him. Every step of the ritual is comforting. If only he didn’t have to cycle the coffee to maintain positive energy levels. He makes a note on his phone, follow up on the latest coffee research, deep dive into timing and “chronographs”.

This article is the first in a two part series that deals with the underlying concepts and mathematics of Pearlian causal inference. Part 2 will focus on a practical example using the DoWhy library. If you prefer to begin with implementation first, skip to Part 2 when it becomes available then return here for the theory.

Harnessing data for better bridges