Control variates for probability and quantile estimation
Article Abstract:
The control variates method is used to extend and improve previous probability and quantile estimators for finite-horizon simulations. The new estimators involve making use of all available information about certain other random variables in the simulation to generate more accurate estimates of probabilities and quantiles of the random variables. The estimators try to get close to the optimal transformation of control variates to improve their precision or else they present a technique for implicitly estimating the optimal multiplier for the control variates of interest. The simple discrete approximations to the optimal transformation are found to perform robustly. The results of an empirical study applying the estimators are discussed.
Publication Name: Management Science
Subject: Business, general
ISSN: 0025-1909
Year: 1998
User Contributions:
Comment about this article or add new information about this topic:
Control variates for quantile estimation
Article Abstract:
Three point and interval estimators are presented for quantiles using a control variate. Point-estimator performance evaluation is obtained through the mean squared error, variance, and bias. Interval-estimator performance evaluation is obtained through the mean, variance, coefficient of variation of the half-width of the interval, and the likelihood that the qth quantile of y is contained in the interval. The new estimators are more effective than the standard estimator with respect to the point estimator's mean squared error and the confidence interval's length.
Publication Name: Management Science
Subject: Business, general
ISSN: 0025-1909
Year: 1990
User Contributions:
Comment about this article or add new information about this topic:
Multivariate batch means and control variates
Article Abstract:
The estimation of a steady-state multivariate mean vector using the combined methods of control-variate variance-reduction and batch-means output-analysis was examined. Multivariate and univariate estimators were also analyzed to determine if they were influenced by the number of batches and the number of control variates. The study resulted in experiment analysis guidelines suggesting a suitable range of the number of batches to select in relation to the number of control responses and control variates.
Publication Name: Management Science
Subject: Business, general
ISSN: 0025-1909
Year: 1992
User Contributions:
Comment about this article or add new information about this topic:
- Abstracts: California is seen giving phone firms profit incentive to control basic rates. Bell Atlantic finds privacy-law abuses in probe of unit; Bell of Pennsylvania workers gave out phone records improperly, parent says
- Abstracts: Complementarity of prior accounting information: the case of stock dividend announcements. Economic sufficiency and statistical sufficiency in the aggregation of accounting signals
- Abstracts: Digital posts first profit in over a year, citing strength overseas, Alpha sales. Wang Labs posts $146.6 million loss for third quarter
- Abstracts: Hewlett posts 44% rise in net for 3rd quarter; result trailed expectations as orders gain slowed in some overseas areas
- Abstracts: Aggregating point estimates: a flexible modeling approach. Unanimity and compromise among probability forecasters