[R] C/C++/Fortran Rolling Window Regressions

[Jason Foster]

I looked into why the coefficients differ in Gabor's example and it's because the example mistakenly switched the order of x and y. The syntax is roll::roll_lm(x, y), which is the same as fastLm, but the example accidentally switched them. After correcting the example, the coefficients are all equal for lm, fastLm, and roll_lm:

library(xts)
library(RcppArmadillo)
library(roll)

set.seed(1)
z <- xts(matrix(rnorm(10), ncol = 2),
         seq(as.Date("2016-08-15"), by = "day", length.out = 5))
colnames(z) <- c("y", "x")

fastLm <- rollapplyr(z, width = 4,
                    function(x) coef(fastLm(cbind(1, x[ , 2]), x[ , 1])),
                    by.column = FALSE)
colnames(fastLm) <- c("(Intercept)", "x")

lm <- rollapplyr(z, width = 4,
                function(x) coef(lm(y ~ x, data = as.data.frame(x))),
                by.column = FALSE)
colnames(lm) <- c("(Intercept)", "x")

roll_lm <- roll_lm(z[ , 2], z[ , 1], 4)$coefficients

all.equal(lm, fastLm)
# [1] TRUE
all.equal(lm, roll_lm)
# [1] TRUE

	[[alternative HTML version deleted]]