hetervar: Heteroscedastic variance estimation for remotely-sensed data

In geoTS: Methods for Handling and Analyzing Time Series of Satellite Images

Heteroscedastic variance estimation for remotely-sensed data

Description

Variance of some remotely-sensed Earth data is time-varying. Utilizing the observations per period (season, year), this function allows for estimation of variability in data either as numeric vector or matricial form

Usage

hetervar(
  x,
  m = NULL,
  lenPeriod = 23,
  method = c("standard", "robust-mad", "robust-Qn")
)

Arguments

x	numeric vector
m	matrix with nrow equal to the number of periods (seasons or years) analyzed and ncol equal to the number of observations per period
lenPeriod	numeric giving the number of observations per period. Default, 23.
method	character specifying whether standard variance, the median absolute deviation (robust-mad) or the more efficient robust variance estimator (robust-Qn) should be used

Details

Designed for data extracted from time series of satellite imagery. Then, it is expected that length(x) be a multiple of lenPeriod. When m is provided, ncol(m) must be equal to lenPeriod. Default of lenPeriod corresponds to the temporal resolution of some MODIS products.

Method standard invokes sd whereas robust-mad uses the median absolute deviation of mad and robust-Qn utilizes the robust scale estimator implemented in Qn.

This function does not allow missing values.

Value

A numeric vector of length lenPeriod

See Also

sd, mad, Qn

geoTS documentation built on Nov. 18, 2022, 1:08 a.m.