stl.control(ss.jump = NULL, inner = NULL, outer = 4, l.degree = 1,
     l.window = NULL, l.jump = NULL, t.degree = 1, t.window = NULL,
     t.jump = NULL, surface = c("interpolate", "direct"), cell =
     0.2, ...)

ss.jump:

skipping value to increase speed of seasonal subseries smoothing in phase one. The subseries smoother skips ahead nsjump points and then linearly interpolates in between. The value should be a positive integer; if it is 1, a smooth is calculated directly at all points. A reasonable choice for ss.jump is 10%-20% of ss.window.

inner:

number of backfitting steps in the inner loop of the phase-one fitting of the seasonal and trend components. The trend component is computed simply to facilitate the proper determination of the seasonal component and is not returned. See the next argument for advice on a value for this argument.

outer:

number of robustness iterations in the outer loop of the phase-one fitting of the seasonal and trend components. The inner loop is nested in these iterations. If the data are well behaved without outliers, then robustness iterations are not needed. In this case set outer to 0, and set inner to 2; the convergence of the inner loop is usually very fast. If outliers are present, setting outer to 4 and inner to 2 usually provides convergence; for near-complete security, set outer to 10.

l.window:

length of window in the loess smoothing portion of the low-pass filter in phase one. The smallest odd integer greater than or equal to frequency(time.series) is the recommended value since it prevents competition between the trend and seasonal components in phase-one fitting. If not an odd integer its given value is increased to the next odd integer.

l.degree:

degree of locally-fitted polynomial in the loess smoothing of the low-pass filter. Values of 0 and 1 are allowed, with the latter recommended for most applications.

l.jump:

skipping value to increase speed of low-pass filter.

t.window:

length of window in the loess trend smoothing in phase one. The smallest odd integer greater than or equal to (1.5 * period)/(1 - 1.5/ss.window) is the recommended value since it prevents competition between the trend and seasonal components in phase-one fitting. If not an odd integer its given value is increased to the next odd integer.

t.degree:

degree of locally-fitted polynomial in the loess trend smoothing in phase one. Values of 0 and 1 are allowed, with the latter recommended for most applications.

t.jump:

skipping value to increase speed of trend smoother.

surface:

loess computation parameter for phase two. See documentation for loess.control.

cell:

loess computation parameter for phase two. See documentation for loess.control.