I have data containing an ordered factor variable with many levels, such as this:
set.seed(1234)
y <- runif(100,0,100)
x <- rep_len(as.character(c(1991:2013)), length.out = 100)
df<-data.frame("x" = factor(x,ordered = TRUE), y)
When I use these data in lm(), R changes the names of estimated coefficients in a way that I think sacrifices clarity.
summary(lm(data = df, formula = y~x))
produces the following:
...Truncated...
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 43.8665 2.9456 14.892 <2e-16 ***
x.L 0.5284 13.9151 0.038 0.970
x.Q -17.6699 14.1375 -1.250 0.215
x.C 0.3310 13.9882 0.024 0.981
x^4 -0.8420 14.0647 -0.060 0.952
x^5 20.1605 14.0629 1.434 0.156
...Truncated...
In order to avoid this, I could "un-order" the factor but this is inconvenient to my circumstance. Is there a way to force R to use the actual level names? Also, what is the explanation of x.L, x.Q and x.C for the first three estimate names, only?
It appears from lack of answers that the best way to accomplish these goals is simply to "un-order" the factor variable, run the linear model and then "re-order" the factor variable, such as:
df$x <- factor(df$x, ordered = FALSE)
m <- lm(data = df, formula = y~x)
df$x <- factor(df$x, ordered = TRUE)