Seasonal Decomposition of Time Series by Loess
Decompose a time series into seasonal, trend and irregular components
loess, acronym STL.
stl(x, s.window, s.degree = 0, t.window = NULL, t.degree = 1, l.window = nextodd(period), l.degree = t.degree, s.jump = ceiling(s.window/10), t.jump = ceiling(t.window/10), l.jump = ceiling(l.window/10), robust = FALSE, inner = if(robust) 1 else 2, outer = if(robust) 15 else 0, na.action = na.fail)
- univariate time series to be decomposed.
This should be an object of class
"ts"with a frequency greater than one.
- either the character string
"periodic"or the span (in lags) of the loess window for seasonal extraction, which should be odd and at least 7, according to Cleveland et al. This has no default.
- degree of locally-fitted polynomial in seasonal extraction. Should be zero or one.
- the span (in lags) of the loess window for trend
extraction, which should be odd. If
NULL, the default,
nextodd(ceiling((1.5*period) / (1-(1.5/s.window)))), is taken.
- degree of locally-fitted polynomial in trend extraction. Should be zero or one.
- the span (in lags) of the loess window of the low-pass
filter used for each subseries. Defaults to the smallest odd
integer greater than or equal to
frequency(x)which is recommended since it prevents competition between the trend and seasonal components. If not an odd integer its given value is increased to the next odd one.
- degree of locally-fitted polynomial for the subseries low-pass filter. Must be 0 or 1.
- s.jump, t.jump, l.jump
- integers at least one to increase speed of
the respective smoother. Linear interpolation happens between every
- logical indicating if robust fitting be used in the
- integer; the number of
inner(backfitting) iterations; usually very few (2) iterations suffice.
- integer; the number of
- action on missing values.
The seasonal component is found by loess smoothing the
seasonal sub-series (the series of all January values, ...); if
s.window = "periodic" smoothing is effectively replaced by
taking the mean. The seasonal values are removed, and the remainder
smoothed to find the trend. The overall level is removed from the
seasonal component and added to the trend component. This process is
iterated a few times. The
remainder component is the
residuals from the seasonal plus trend fit.
Several methods for the resulting class
"stl" objects, see,
stlreturns an object of class
time.series a multiple time series with columns
weights the final robust weights (all one if fitting is not done robustly). call the matched call. win integer (length 3 vector) with the spans used for the
deg integer (length 3) vector with the polynomial degrees for these smoothers. jump integer (length 3) vector with the jumps(skips) used for these smoothers. ni number of inner iterations no number of outer robustness iterations
This is similar to but not identical to the
stl function in
remainder component given by S-PLUS is the sum of
remainder series from this function.
R. B. Cleveland, W. S. Cleveland, J.E. McRae, and I. Terpenning (1990) STL: A Seasonal-Trend Decomposition Procedure Based on Loess. Journal of Official Statistics, 6, 3--73.
StructTS for different kind of decomposition.
require(graphics) plot(stl(nottem, "per")) plot(stl(nottem, s.window = 7, t.window = 50, t.jump = 1)) plot(stllc <- stl(log(co2), s.window = 21)) summary(stllc) ## linear trend, strict period. plot(stl(log(co2), s.window = "per", t.window = 1000)) ## Two STL plotted side by side : stmd <- stl(mdeaths, s.window = "per") # non-robust summary(stmR <- stl(mdeaths, s.window = "per", robust = TRUE)) op <- par(mar = c(0, 4, 0, 3), oma = c(5, 0, 4, 0), mfcol = c(4, 2)) plot(stmd, set.pars = NULL, labels = NULL, main = "stl(mdeaths, s.w = "per", robust = FALSE / TRUE )") plot(stmR, set.pars = NULL) # mark the 'outliers' : (iO <- which(stmR $ weights < 1e-8)) # 10 were considered outliers sts <- stmR$time.series points(time(sts)[iO], 0.8* sts[,"remainder"][iO], pch = 4, col = "red") par(op) # reset