`stl()` function

> stl(x, s.window = "periodic")
> # additional arguments removed

with the two mandatory arguments x, for the time series data, and s.window chosen as periodic assuming that the seasonal pattern is constant through time. On the help page of the function is mentioned that x should be a ts object.

More about the theory behind this function can be found here.

As an example we analyze monthly birth rates data for France from the National Institute of Statistics and Economic Studies (France) database.

> births <- Quandl("INSEE/62_001641602_M", type = "ts")
> plot(births)

plot of chunk birthratefrance

It is obvious that the birth rates time series exhibits a seasonal pattern and a trend component.

> births_stl <- stl(births, s.window = "periodic")
> # s.window = "periodic" means that we assume the seasonal pattern
> # is constant through time
> 
> plot(births_stl)

plot of chunk salesstl

stl() function returns a matrix with the seasonal, trend and remainder component.

> stl_mat <- births_stl$time.series
> head(stl_mat)

         seasonal    trend   remainder
[1,] -0.253229421 12.45112 -0.49789058
[2,] -0.233778895 12.46242 -0.12864439
[3,] -0.423419212 12.47373  0.04969264
[4,] -0.235228039 12.48372  0.25150890
[5,]  0.002962996 12.49371  0.50332530
[6,]  0.106144514 12.50664  0.38721940

stl() function

`stl()` function