Parameter Estimators for Gaussian Models with Censored Time Series and Spatio-temporal Data

Computationally-fast algorithms are considered for estimating parameters in Gaussian time series and spatio-temporal models from censored and/or missing data. The problem arises in fitting models involving Gaussian latent variables to environmental data. Spectral estimators and least-squares fits of auto- and cross-covariances are found to be of similar efficiency for fitting models to rainfall and solar radiation data.