step4.R

## We're going to be running regressions...
## If a predicted value is positive, we're going to say it's a prediction for hamilton authorship.
## If a predicted value is negative, we're going to say it's a prediction for madison authorship.

author <- rep(NA, nrow(dtm1)) # a vector with a missing value
author[hamilton] <- 1 # 1 if Hamilton
author[madison] <- -1 # -1 if Madison

## data frame for regression
author.data <- data.frame(author = author[c(hamilton, madison)],
                          tfm[c(hamilton, madison), ])

## To predict the authorship, we use the term frequency of the 4 words 
## selected based on our preliminary aanalysis ('upon', 'there', 
## 'consequently', and 'whilst'. The data frame object we created above contains
## the term frequency of the 10 words including these 4. We estimate the 
## coefficients using the 66 essays with known authorship.

hm.fit <- lm(author ~ enough + upon + although + whilst + always + commonly
                      + consequently + considerable, data = author.data)
hm.fit