I have a data set with 358 rows, a quick look is as follows
> unempts gdpts
> [1,] 7.03 4.2
> [2,] 7.17 3.7
> [3,] 6.97 3.2
> [4,] 6.83 2.9
> [5,] 6.60 2.7
> [6,] 6.27 3.4
I need to run a regression of Column 1 on Column 2, for 60 rolling windows and save all results in matrix so that I can plot the same and run HAC.
I tried the following:
regRO <- rollapply(datam, 60, lm, by.column = TRUE,fill=NA)
but get an error as follows
Error in formula.default(object, env = baseenv()) : invalid formula
Please advise what needs to be changed.
You have to specify the function call explicitly so that X is first converted to a data frame to satisfy the input requirements of lm:
library(zoo)
set.seed(123) # ensure runif in next line uses same random numbers each time run
datam <-matrix(runif(200),ncol=2)
out <- rollapplyr(datam,
width = 60,
FUN = function(X) coef(lm(as.data.frame(X))),
by.column=FALSE,
fill = NA) # omit fill=NA if NA rows not wanted
giving:
> tail(out)
(Intercept) V2
[95,] 0.5071618 -0.1041905
[96,] 0.5015140 -0.1032419
[97,] 0.5196683 -0.1404855
[98,] 0.5171599 -0.1401534
[99,] 0.5122871 -0.1288700
[100,] 0.5176396 -0.1297019