Writing
Simulations, mostly. I read the paper or the documentation, generate data where I control the ground truth, run the method, and write up what actually happened, including when the result is inconvenient.
Speed vs. Power: The Hidden Tradeoffs of Sequential Testing
Sequential testing fixes the peeking problem, but the safety net costs you power. Here is exactly how much, across 10,000 simulations of O'Brien-Fleming against Pocock.
Read the write-up →
Beyond the Peek: Using Sequential Boundaries to Protect Experiment Integrity
Peeking daily over a 30-day experiment inflates your false positive rate from 5% to 26%. Group sequential boundaries fix it, and choosing between O'Brien-Fleming and Pocock depends on what you are optimizing for.
Read the write-up →
Real-world Data Science Problem: Synthetic Controls using Causal Impact — BEWARE
Synthetic control methods are a core technique for data scientists specializing in causal inference methods. One of the most popular packages to estimate synthetic controls is the CausalImpact package by Google. In this real…
Read the write-up →
Data Science Tutorial: The Event Study — A powerful causal inference model
Here’s a short tutorial and example of an Event Study, a popular and flexible causal inference model. Event study models can be used for a range of business problems including estimating: Excess stock price…
Read the write-up →
Google’s Marketing Mix Model: Dramatically improved MMM results from Meridian
In my previous video, I introduced Google’s marketing mix model called Meridian which was just publicly released open source this past week. In that video, I demonstrated how naive usage of prepackaged models can…
Read the write-up →
Google Meridian: Marketing Mix Model can give crazy results if you aren’t careful
I tested Google’s new Marketing Mix Model aka Meridian with simulated data and it demonstrates how dangerous data science models can be in the wrong hands. Video breakdown here: https://youtu.be/wQbk2TWaH50 I simulated data so…
Read the write-up →
Big news for data scientists: Google released its Meridian Marketing Mix Model for FREE
Data scientists regularly have to demonstrate impact: How well does your work justify your current and future compensation? Large companies may have advertising budgets in the hundreds of millions of dollars so that even…
Read the write-up →
Building a Networking Feedback Loop for Career Success
Want to turn cold outreach into warm referrals, land research opportunities, and connect with industry professionals—all while building lasting relationships? Here’s how you can create a powerful, sustainable networking feedback loop: Start with alumni:…
Read the write-up →
How your interview is like a real-world data science problem
How are your interviews like real-world data science problems? 🤔 In predictive models, the most important feature is often whether the user has already done the action you’re trying to predict. Now put yourself…
Read the write-up →
LinkedIn networking: How I turned messages into job offers
Each of my last two jobs have come through LinkedIn, and I’ve had multiple offers that were initiated from LinkedIn messages. From my experience both sending and receiving messages, here are the sweet spots…
Read the write-up →