Spatial Estimation and Prediction for Censored/Missing Responses
It provides functions to estimate the parameters in spatial models with censored/missing responses via the Expectation-Maximization (EM) algorithm (see Dempster, Laird, and Rubin (1977)< https://www.jstor.org/stable/2984875>), the Stochastic Approximation EM (SAEM) algorithm (see Delyon, Lavielle, and Moulines (1999)< https://www.jstor.org/stable/120120>), and the Monte Carlo EM (MCEM) algorithm (see Wei and Tanner (1990)<10.1080>). These algorithms are widely used to compute the maximum likelihood (ML) estimates in incomplete data problems. The EM algorithm computes the ML estimates when a closed expression for the conditional expectation of the complete-data log-likelihood function is available. In the MCEM algorithm, the conditional expectation is substituted by a Monte Carlo approximation based on many independent simulations of the missing data, while the SAEM algorithm splits the E-step into a simulation step and an integration step. The SAEM algorithm was developed as an alternative to the computationally intensive MCEM algorithm.
This package also approximates the standard error of the estimates using the method developed by Louis (1982)< https://www.jstor.org/stable/2345828>. It also has a function that performs spatial prediction in a set of new locations. Besides the functions to estimate parameters, this package allows computing the covariance matrix and the distance matrix.10.1080>