Through Monte Carlo procedures, three different techniques for estimating the parameter theta (proportion of the "shocks" remaining in the system) in the Integrated Moving Average (0,1,1) time-series model are compared in terms of (1) the accuracy of the estimates, (2) the independence of the estimates from the true value of theta, and (3) the independence of the estimates from a 'shift in level' in the time-series following an intervention. In the "usual" range for theta, the methods appear equally accurate. One produces complex estimates in special cases. Estimates are independent of the true value and changes in level. (Author)