From the course: Becoming a Good Data Science Customer
Join today to access over 25,500 courses taught by industry experts.
Testing hypothesis: Effect sizes and p-values
From the course: Becoming a Good Data Science Customer
Testing hypothesis: Effect sizes and p-values
- [Narrator] Exploratory data analyses are just that, exploratory. They are meant to generate questions and hypotheses that can be more thoroughly explored in subsequent analyses. Agreeing on what hypothesis should be tested is an important step in the modeling process. Data science customers may not think about hypothesis testing, but they probably should. It's important to understand what questions cannot be answered with a statistical hypothesis test. For example, hypothesis tests are often not useful for why questions. Hypothesis tests typically have two outputs, an effect size and a P-value. If we are comparing two models, the effect size would be a measure of how much better the new model was at detecting fraud, increasing response or whatever was the outcome of interest. "Is the effect size big enough?" is a domain question. Sometimes any improvement is what we're looking for. While other times we want the improvement to be at least a certain percent better. The effect estimate…
Practice while you learn with exercise files
Download the files the instructor uses to teach the course. Follow along and learn by watching, listening and practicing.
Download courses and learn on the go
Watch courses on your mobile device without an internet connection. Download courses using your iOS or Android LinkedIn Learning app.