Perform an STL like (based on Loess) decomposition on any periodicity
Usage
stlplus(
y,
period,
multiplicative = TRUE,
swindow = 7,
twindow = 0,
lwindow = 0,
sdegree = 0,
tdegree = 1,
ldegree = 1,
sjump = 0,
tjump = 0,
ljump = 0,
ninnerloop = 1,
nouterloop = 15,
weight.threshold = 0.001,
weight.function = c("BIWEIGHT", "UNIFORM", "TRIANGULAR", "EPANECHNIKOV", "TRICUBE",
"TRIWEIGHT"),
legacy = FALSE
)Arguments
- y
input time series.
- period
period, any positive real number.
- multiplicative
Boolean indicating if the decomposition mode is multiplicative (TRUE).
- swindow
length of the seasonal filter.
- twindow
length of the trend filter.
- lwindow
length of the filter used to remove the trend of the seasonal
- sdegree
degree of the seasonal local polynomial (0 or 1)
- tdegree
degree of the trend local polynomial (0 or 1)
- ldegree
degree of the low-pass local polynomial (0 or 1)
- sjump
number of jumps in the computation of the seasonal
- tjump
number of jumps in the computation of the trend
- ljump
number of jumps in the computation of the trend in the seasonal
- ninnerloop
Number of inner loops
- nouterloop
Number of outer loops (computation of robust weights)
- weight.threshold
Weights threshold (in [0, 0.3])
- weight.function
weights function
- legacy
use of the (bugged) legacy MAD