rtaph
Coursera Class (Data Products)
This presentation demonstrates that simple linear regression methods are acceptable even for complex sampling designs, where the assumptions of a simple random sample are violated.
Visual comparison of both regression methods (simple and weighted) on a real data set. The user can mix-and match variables as he/she pleases and see that both methods produce nearly identical results.
The plot on the following slide shows weighted data from the California Academic Performance Index (API). The data set follows a stratified sampling design. Dynamic univariate models are presented for both weighted and normal regression approaches, in the usual form
\[ \hat{Y} = \beta_0 + \beta_1 X_i + \epsilon \]
The user can select any two variables (out of 25 total). The output should demonstrate that the weighted model (in red) is almost identical to the regular regression model (in blue). The exact regression equation for both methods is provided to the left of the visualization.