Search

Two models are presented for the response times of different operators to different tasks where response is initiated by one or more cues provided by the system. One model for the log-response times is a mixed or latent factor model with unequal case fixed effects and variances. The other model for the log-response times is a non-Gaussian log-extreme-value model. Procedures for estimating the parameters by maximum likelihood are presented. The models are used to analyze response time data from simulator experiments involving nuclear power plant operators performing certain safety-related tasks. The findings of the models are critiqued and applications to risk analysis are sketched. (KR)

This paper studies prediction of future failure (rates) by hierarchical empirical Bayes (EB) Poisson regression methodologies. Both a gamma distributed super-population as well as a more robust (long-tailed) log student-t super-population are considered. Simulation results are reported concerning predicted Poisson rates. The results tentatively suggest that a hierarchical model with gamma super-population can effectively adapt to data coming from a log-Student-t-super-population particularly if the additional computation involved with estimation for the log-Student-t hierarchical model is burdensome.

Two models are presented for the response times of different operators to different tasks where response is initiated by one or more cues provided by the system. One model for the log-response times is a mixed or latent factor model with unequal case fixed effects and variances. The other model for the log-response times is a non-Gaussian log-extreme-value model. Procedures for estimating the parameters by maximum likelihood are presented. The models are used to analyze response time data from simulator experiments involving nuclear power plant operators performing certain safety-related tasks. The findings of the models are critiqued and applications to risk analysis are sketched. Keyword: Extreme-value distribution; Weibull distribution. (KR)

This paper studies prediction of future failure (rates) by hierarchical empirical Bayes (EB) Poisson regression methodologies. Both a gamma distributed super-population as well as a more robust (long-tailed) log student- t super-population are considered. Simulation results are reported concerning predicted Poisson rates. The results tentatively suggest that a hierarchical model with gamma super-population can effectively adapt to data coming from a log-Student-t-super-population particularly if the additional computation involved with estimation for the log-Student-t hierarchical model is burdensome, Naval Postgraduate School Research Council Research Program., http://archive.org/details/regressionanalys00gave, O&MN, Direct Funding

Several (I in number) observers cooperate to identify the member of a population of items (J in number) transmitting. The observers' measurements of basic item-identifying (known) parameters are error-prone; errors are frequently non-Gaussian (gross-error-prone). Various methods are introduced and studied for combining the observers; observations so as to achieve as high a probability of correct identification as possible. Linear combinations of observations are shown to be less than satisfactory in the environment described. Continuation and generalization of the work is in progress. Keywords: robust estimation in a restricted parameter space; Cauchy and contaminated normal measurement errors; maximum likelihood and Bayes estimates; the discrimination problem.

In this work, several (5) statistical estimators for solving the typical remote sensing 'inverse' problem are explained, illustrated and evaluated using both real data sets connecting surface windspeed and whitecap coverage and simulated data. The results suggests the superiority of some of these estimators.

This article addresses the problem of predicting multi(more than two) category events. It describes a new approach (based on generalized linear models) which is natural extension of the technique of logistic regression (a method widely used for the prediction of dichotomous events). As an example, the method is applied to the prediction of tropical storm imminence, and the performance of the method is compared with that of alternative well-established techniques. Keywords: Meteorology; Weather forecasting.

In this work, several (5) statistical estimators for solving the typical remote sensing 'inverse' problem are explained, illustrated and evaluated using both real data sets connecting surface windspeed and whitecap coverage and simulated data. The results suggests the superiority of some of these estimators, Prepared for: Office of Naval Research, Arlington, VA., http://archive.org/details/estimationofseas00omui, N0001485WR24061, NA

A number, I, of nominally similar items generate events (e.g. failures) at possibly different rates, or mean time intervals. This paper addresses the problem of appropriately pooling the data from the different sources. The approach is parametric empirical Bayes: true individual item rates are assumed to come from a fixed superpopulation. It is shown how parameters of a superpopulation model can be estimated from all of the data, and combined with individual unit history, can provide improved estimates of individual rates. The procedure can be robust: evidence that a particular rate is far off from the main body of rates permits that outlier to stand by itself, i.e. to resist pooling. Illustrative analyses of data are supplied. Keywords: Robustness; Population(Mathematics); and Charts, Prepared for: Naval Postgraduate School, http://archive.org/details/robustempiricalb00gave

This article addresses the problem of predicting multi(more than two) category events. It describes a new approach (based on generalized linear models) which is natural extension of the technique of logistic regression (a method widely used for the prediction of dichotomous events). As an example, the method is applied to the prediction of tropical storm imminence, and the performance of the method is compared with that of alternative well-established techniques. Keywords: Meteorology; Weather forecasting, Naval Postgraduate School, Monterey, CA., http://archive.org/details/predictionofpoly00omui

Several (I in number) observers cooperate to identify the member of a population of items (J in number) transmitting. The observers' measurements of basic item-identifying (known) parameters are error-prone; errors are frequently non-Gaussian (gross-error-prone). Various methods are introduced and studied for combining the observers; observations so as to achieve as high a probability of correct identification as possible. Linear combinations of observations are shown to be less than satisfactory in the environment described. Continuation and generalization of the work is in progress. Keywords: robust estimation in a restricted parameter space; Cauchy and contaminated normal measurement errors; maximum likelihood and Bayes estimates; the discrimination problem, National Security Agency, Fort Meade, MD and the Office of Naval Research, http://archive.org/details/problemsofidenti00gave