Exercise

Toying with OLS II: Statistical Power

Since OLS models are so easy to visualize, they provide a nice tool for understanding why models that use more observations tend to have more statistical power. In the following question, we have two datasets provided by the popular online auctioneer eGulf about the relationship between WePhone age and sales price, one with 25 observations. For each of these datasets, we add 25 randomly generated observations that have no relationship between WePhone age and sales price, to represent "statistical noise" - unexplained variation in our sample. Follow the sample code to create regression models with these two datasets, and answer the following question:

Instructions

100 XP
  • 1) Take a look at summary statistics of the two WePhone datasets.
  • 2) Build glm models that estimate the effect of phone ages on sales prices in both datasets.
  • 3) Which regression model most accurately represents the true relationship between WePhone Age and SalesPrice?