Created
July 31, 2012 19:32
-
-
Save mmparker/3219819 to your computer and use it in GitHub Desktop.
Regression with a group variable and a group-trait variable
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
# Make a dataset for Channel A | |
dat.a <- data.frame(purchase = rbinom(n = 100, size = 1, prob = .1), | |
channel = "a") | |
# One for Channel B | |
dat.b <- data.frame(purchase= rbinom(n = 100, size = 1, prob = .2), | |
channel = "b") | |
# Add the rates | |
dat.a$rate <- sum(dat.a$purchase) / nrow(dat.a) | |
dat.b$rate <- sum(dat.b$purchase) / nrow(dat.b) | |
# Merge into one dataset | |
dat <- rbind(dat.a, dat.b) | |
# Model purchase as a function of channel, rate, and both | |
mchannel <- glm(purchase ~ channel, data = dat) | |
mrate <- glm(purchase ~ rate, data = dat) | |
mboth <- glm(purchase ~ channel + rate, data = dat) | |
# Predictions from the models are all the same, | |
# but R will complain that "prediction from a rank-deficient fit may be misleading" | |
data.frame(mchannel = predict(mchannel, dat), | |
mrate = predict(mrate, dat), | |
mboth = predict(mboth, dat) | |
) | |
# Which is because when both of those variables are in the model, | |
# R recognizes that there is zero new information in the rates | |
# and doesn't estimate any coefficients for them | |
summary(mboth) |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment