Skip to content

Instantly share code, notes, and snippets.

@carlislerainey
Created December 1, 2017 23:05
Show Gist options
  • Save carlislerainey/7cc726c27982d674082da94d86f04306 to your computer and use it in GitHub Desktop.
Save carlislerainey/7cc726c27982d674082da94d86f04306 to your computer and use it in GitHub Desktop.
code to understand product terms and probit models
# code to understand product terms and probit models
# clear workspace
rm(list = ls())
# load packages
library(tidyverse)
library(plotly)
# parameters
n_pts <- 100 # number of values of x (and z) at which we calcalte Pr(y = 1)
b_cons <- -3 # Pr(y) = Phi(b_cons + b_x*x + b_z*z + b_xz*x*z)
b_x <- 0
b_z <- -3
b_xz <- 6
# create data for surface
x_pred <- z_pred <- seq(0, 1, length.out = n_pts) # values at which to calculate Pr(y)
df <- data.frame(x = x_pred, z = z_pred) %>% # put values in data frame
expand(x, z) %>% # expand data frame to all combinations
mutate(prob = pnorm(b_cons + b_x*x + b_z*z + b_xz*x*z)) # calculate Pr(y) for all combinations
# put Pr(y) or df$prob in a numeric matrix
X <- matrix(df$prob, nrow = n_pts, ncol = n_pts, byrow = FALSE)
# plot surface
plot_ly() %>%
add_surface(x = x_pred, y = z_pred, z = X) %>%
layout(scene = list(xaxis = list(title = "x"),
yaxis = list(title = "z"),
zaxis = list(title = "Probability")))
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment