spacetime.control: Set smoothing parameters for the drsstl fitting

In XiaosuTong/drSpaceTime: Spatial Seasonal Trend Loess analysis routine using divide and recombined

Description

Set control parameters for the smoothing fit of stl and spatial smoothing

Usage

spacetime.control(vari = "resp", time = "date", n, stat_n, n.p = 12,
  s.window, s.degree = 1, t.window = NULL, t.degree = 1, inner = 2,
  outer = 1, mthbytime = 1, s.jump = 10, t.jump = 10, cell = 0.2,
  degree, span, Edeg, surf = c("direct", "interpolate"),
  family = c("symmetric", "gaussian"), siter = 2)

Arguments

vari	variable name in string of the response variable. The default is "resp"
time	variable name in string of time index of the whole time series. The default is "date". In the final results on HDFS, the index of time will be changed to this variable instead of year and month.
n	the number of total observations in the time series at each location.
stat_n	The number of stations.
n.p	the number of observations in each subseries. It should be 12 for monthly data for example.
s.window	either the character string "periodic" or the span (in lags) of the loess window for seasonal extraction, which should be odd. This has no default.
s.degree	degree of locally-fitted polynomial in seasonal extraction. Should be 0, 1, or 2.
t.window	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.
t.degree	degree of locally-fitted polynomial in trend extraction. Should be 0, 1, or 2.
inner	The iteration time for inner loop of stlplus for time dimension fitting
outer	The iteration time for outer loop of stlplus for time dimension fitting
mthbytime	The number of months will be grouped together in the by time division after swaptoTime. The parameter is only used for swaptoTime. Since there may be to many time point in each location, the swaptoTime looping all time point will add to much overhead caused by rhcollect. So every mthbytime time point are collect into one key-value pair. It is save to leave it as default 1. If the time series is extremely long, it can be set to be 2.
s.jump, t.jump	integers at least one to increase speed of the respective smoother. Linear interpolation happens between every '*.jump'th value.
cell	if interpolation is used this controls the accuracy of the approximation via the maximum number of points in a cell in the kd-tree. Cells with more than 'floor(n*span*cell)' points are subdivided.
degree	smoothing degree for the spatial loess smoothing. It can be 0, 1, or 2.
span	smoothing span for the spatial loess smoothing.
Edeg	the degree for the conditioanl parametric model including elevation.
surf	should the fitted surface be computed exactly or via interpolation from a kd tree.
family	if '"gaussian"' fitting is by least-squares, and if '"symmetric"' a re-descending M estimator is used with Tukey's biweight function.
siter	the number of iterations used for the spatial smoothing procedure if family is set to be '"symmetric"', which is for robust fitting.

Value

A list with space-time fitting parameters.

Author(s)

Xiaosu Tong

References

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.

Examples

    spacetime.control(
     n = 576, stat_n = 7738, n.p = 12, s.window = 21, s.degree = 1, t.window = 241,
      t.degree = 1, degree = 2, span = 0.015, Edeg = 2, surf = "interpolate"
    )

XiaosuTong/drSpaceTime documentation built on May 9, 2019, 11:06 p.m.