inst/shiny-apps/interactions/interactions_help.md

In clayford/consultr: Shiny Apps for Consulting and Tutoring

When the app is initialized, a random data set of predictors is loaded. X1 and X2 are each 100 random draws from an N(5,1) distribution. X1 and X2 do not change while the app is in use. Each time a slider is moved, a new response vector (y) is generated using the predictors and their associated coefficients as follows:

y = β₀ + β₁X1 + β₂X2 + β₃X1X2 + N(0,σ)

Think of the slider values as the “true” model and the summary output as the attempt to recover the true values.

The plot shows the relationship between X1 and y at four different values of X2. When there is no interaction (i.e., β₃ = 0), the lines should all have roughly the same slope. In that case, it doesn’t matter what value X2 takes; the relationship between X1 and y remains the same. However, when there is a non-zero interaction, the relationship between X1 and y depends on X2. Adjusting the sliders allows one to explore how the relationship between X1 and y changes given various magnitudes and directions (positive or negative) of the coefficients. See if you can guess in what ways the relationships will change as you adjust the coefficients.

clayford/consultr documentation built on Aug. 5, 2021, 7:29 p.m.