unit_09_part_3.md

Unit 9: Part 3

In this section, which we anticipate might take about 90 minutes to complete, you will learn about how to reason about the sampling uncertainty in estimates produced from an OLS regression.

Recall that when we read about sampling, in an earlier section, we used convergence in distribution from the Central Limit Theorem to make statements about test statistics. In this section, we will discuss the appropriate test statistics for three levels of inference:

1. The internal model weights, where you will be able to provide a statistically sound answer to the question, "Is this coefficient significantly from the null hypothesis?"
2. The overall model, where you will be able to provide a statistically sound answer to the question, "Is this model significantly different from the null model?"
3. The output from the model, namely the model predictions, where you will be able to provide a statistically sound answer to the question, "What is the most likely value that this model predicts, given input data? How different from this most likely value could a prediction be?"