Patching rainfall data using regression methods .2. Comparisons of accuracy, bias and efficiency

Six methods of patching rainfall data via regression methods were proposed in a companion paper, three using methods of best subset selection and three using methods based on the Expectation-Maximization (EM) algorithm. The first set is specifically designed to yield good estimates of the data missing at a target site when there are data missing in the control sites also. The second set can accommodate a reasonable proportion of missing data in all the sites and proceed to patch all the missing data using iterative computation procedures. In this, the second in a set of three papers, comparisons between these methods based on accuracy, variance preservation and speed are made and it turns out that the simplest method, the pseudo-EM Method 2, is best on all these counts.