vetools (version 1.3-28)

complete.series: Complete relatively large holes in data-sets

Description

This functions completes relatively large holes in monthly time-series objects.

Usage

complete.series(collection, model, k.ubic = NA, centers = 3, nstart = 3, weps = 0.05, MAX.ITER = 100, AEM.debug = T)

Arguments

collection
A list of class Catalog that contains the objects to complete.
model
A list of fixed-effects models related to collection$data.
k.ubic
A data.frame of exactly one member k.ubic$cluster which is a scalar vector of length equal to collection$data and specifying to which cluster belongs to each element of the list collection$data.
centers
If k.ubic is unavailable, this sets the quantity of clusters to build.
nstart
If k.ubic is unavailable, then this parametre sets the initial quantity of center with which to start the k-means algorithm.
weps
Tolerance for the E-M Algorithm.
MAX.ITER
Maximum number of iterations for the E-M Algorithm.
AEM.debug
Logical flag indicating if verbosity is required.

Value

Returns a completed version of collection (collection$data).

Details

The main idea behind this functions is to complete the time-series of the list by first clustering similar stations and then applying to each cluster the E-M Algorithm in order to complete the series. The E-M Algorithms is an iterative method that in each iteration performs two tasks: fist estimates the expected values and then maximizes their likelyhood. This goes on util some stopping criteria is meat.

See Also

fill.small.missing