SERVO: Simulation experiments with random-vector output

The authors discuss statistical methods, system design, and examples for SERVO, a software package that supports simulation experiments having output data that are assumed to be a sequence of identically distributed random vectors. SERVO requires two types of input: a description of the model of interest and a description of the desired experiment. From this input, SERVO generates C-language code that when executed is the simulation experiment. The experiment automatically creates an output report with point estimators and associated standard errors for a variety of typical performance measures. The marginal performance measures include means, standard deviations, variances, mean squared errors, third and fourth standardized moments, and points on the distribution function; the joint performance measures are covariances and correlations.<
>