Categorical numerical variable into continuous form for regression problem

Question

I have a dataset in which all columns are numerical.Some of the columns have categories in numerical form having levels >= 2. Do i need to convert that categorical numerical column into factor for regression analysis or not ? Please suggest any better approach in R.

Answer 1

Yes you do. You can prove it to yourself...

x <- rep(1:5, 20)
y <- rnorm(100)

# not converting to factors
m1 <- lm (y ~ x)

# converting to factors
m2 <- lm(y ~ as.factor(x) )

summary(m1) # one fitted coefficent
summary(m2) # five fitted coefficients