Search NASA⌕ Search

Engineering topics

Hamada, Michael S.

Publications and source records attributed to Hamada, Michael S..

Analysis of overlapping count data

Counts of a specific characteristic were obtained within regions defined on an object that was manufactured in a proprietary setting. The count regions were altered during production and resulted in misaligned or overlapping count data. A closed-formula maximum likelihood estimator (MLE) of the new region means is derived using all of the available count data and an independent Poisson model. The MLE is shown to be preferable to estimators constructed using generalized linear models for the overlapping data setting. This closed-form estimator extends to over-dispersed overlapping count data as the quasi-MLE and also performs well with correlated overlapping count data. Standard errors for the estimator are approximated and are validated with a simulation study. Additionally, the methods are extended to overlapping multinomial data. Illustrative examples of the methods are provided throughout the paper and are reproducible with the supplemental R code. Additionally, proofs of the paper’s results are also included in the supplemental material.

97 MATHEMATICS AND COMPUTING↗

Analyzing count data with measurement error

In this article, we analyze observed count data such as the number of defects in a steel product where the observed counts are the true counts measured with errors. We account for the measurement error by using a measurement error model based on a latent lognormal (LLN) distribution. We consider making inference about a single population (e.g., from samples of a production lot) and a regression model (e.g., from runs of a designed experiment), where the measurement system properties are known, that is, the parameters of the LLN distribution are known. Then, we consider simultaneous inference for the single population and regression model as well as the measurement system. We demonstrate the proposed methodology with both simulated and real observed counts.

42 ENGINEERING↗

A note on minimizing time assurance tests for repairable systems

We consider assurance testing for repairable systems when supplementary information is available in addition to the data collected in the assurance test. Here, the supplementary information is incorporated using a Bayesian inferential framework. Here we consider assurance testing for a homogeneous Poisson process. In this note we consider an alternative criterion that minimizes the test time while ensuring that the requirements on the producer's and consumer's risks are met. We illustrate the use of this alternative criterion with an example.

42 ENGINEERING↗

A note on the Bayesian analysis of experiments with correlated multiple responses using a matrix-logarithmic covariance model

In the literature, analysis of multiple responses from experiments with replicates has modeled the covariance matrix directly as linear models of the transformed variances and correlations, i.e., covariance modeling. This article considers models based on the matrix- logarithm of the covariance matrix. This so-called log-covariance modeling is illustrated with data from actual experiments and compared with the traditional covariance modeling.

97 MATHEMATICS AND COMPUTING↗

Quality quandaries: Experiencing the practice of statistics with simple experiments

This article illustrates the practice of statistics through simple experiments. In running ex- periments that involve blowing bubbles, collecting rain, and spinning propellers, issues arise that in the practice of statistics require being carefully addressed to correctly apply statisti- cal methods. Addressing these issues involves developing statistical models, simulation, and optimization. Sometimes the solutions require research, i.e., applied research.

42 ENGINEERING↗

Analysis of Repeatability and Reproducibility Studies With Ordinal Measurements

A Bayesian inferential approach with a noninformative prior is introduced to analyze ordinal repeatability and reproducibility (R&R) data using the De Mast–Van Wieringen model. This approach is extended with a weakly informative prior and random effects to allow for the consideration of a population of raters and prediction of a new rater. This random-effects approach is also shown to result in partial pooling of estimates across raters. In addition, match-probability-based measures to decompose ordinal R&R study data into contributions due to repeatability and due to reproducibility are defined. All extensions involving Bayesian inference (for fixed or random effects) and measures are illustrated on real and simulated ordinal R&R study data and are applicable in business and industry settings. This methodology can be implemented using the supplemental R package ordinalRR available from CRAN. Additional supplementary material for this article is available online.

97 MATHEMATICS AND COMPUTING↗

Comparing Methods for Assessing Reliability Uncertainty Based on Pass/Fail Data Collected Over Time

In this paper, we compare statistical methods for analyzing pass/fail data collected over time; some methods are traditional and one (the RADAR or Rationale for Assessing Degradation Arriving at Random) was recently developed. These methods are used to provide uncertainty bounds on reliability. We make observations about the methods' assumptions and properties. Finally, we illustrate the differences between two traditional methods, logistic regression and Weibull failure time analysis, and the RADAR method using a numerical example.

97 MATHEMATICS AND COMPUTING↗

Selecting an Informative/Discriminating Multivariate Response for Inverse Prediction

nverse prediction is important in a wide variety of scientific and engineering contexts. One might use inverse prediction to predict fundamental properties/characteristics of an object using measurements obtained from it. This can be accomplished by “inverting” parameterized forward models that relate the measurements (responses) to the properties/characteristics of interest. Sometimes forward models are science based; but often, forward models are empirically based, using the results of experimentation. For empirically-based forward models, it is important that the experiments provide a sound basis to develop accurate forward models in terms of the properties/characteristics (factors). While nature dictates the causal relationship between factors and responses, experimenters can influence control of the type, accuracy, and precision of forward models that can be constructed via selection of factors, factor levels, and the set of trials that are performed. Whether the forward models are based on science, experiments or both, researchers can influence the ability to perform inverse prediction by selecting informative response variables. By using an errors-in-variables framework for inverse prediction, this paper shows via simple analysis and examples how the capability of a multivariate response (with respect to being informative and discriminating) can vary depending on how well the various responses complement one another over the range of the factor-space of interest. Insights derived from this analysis could be useful for selecting a set of response variables among candidates in cases where the number of response variables that can be acquired is limited by difficulty, expense, and/or availability of material.

42 ENGINEERING↗

Statistical mentoring at early training and career stages

At Los Alamos National Laboratory (LANL), statistical scientists develop solutions for a variety of national security challenges through scientific excellence, typically as members of interdisciplinary teams. At LANL, mentoring is actively encouraged and practiced to develop statistical skills and positive career-building behaviors. Mentoring activities targeted at different career phases from student to junior staff are an important catalyst for both short and long term career development. This article discusses mentoring strategies for undergraduate and graduate students through internships as well as for postdoctoral research associates and junior staff. Topics addressed include project selection, progress, and outcome; intellectual and social activities that complement the student internship experience; key skills/knowledge not typically obtained in academic training; and the impact of such internships on students’ careers. Experiences and strategies from a number of successful mentorships are presented. Feedback from former mentees obtained via a questionnaire is incorporated. As a result, these responses address some of the benefits the respondents received from mentoring, helpful contributions and advice from their mentors, key skills learned, and how mentoring impacted their later careers.

97 MATHEMATICS AND COMPUTING↗

Using Bayesian variable selection to analyze regular resolution IV two-level fractional factorial designs

Regular two-level fractional factorial designs have complete aliasing in which the associated columns of multiple effects are identical. Here, we show how Bayesian variable selection can be used to analyze experiments that use such designs. In addition to sparsity and hierarchy, Bayesian variable selection naturally incorporates heredity . This prior information is used to identify the most likely combinations of active terms. We also demonstrate the method on simulated and real experiments.

42 ENGINEERING↗

Design of experiments and data analysis challenges in calibration for forensics applications

Forensic science aims to infer characteristics of source terms using measured observables. Our focus is on statistical design of experiments and data analysis challenges arising in nuclear forensics. More specifically, we focus on inferring aspects of experimental conditions (of a process to produce product Pu oxide powder), such as temperature, nitric acid concentration, and Pu concentration, using measured features of the product Pu oxide powder. The measured features, Y, include trace chemical concentrations and particle morphology such as particle size and shape of the produced Pu oxide power particles. Making inferences about the nature of inputs X that were used to create nuclear materials having particular characteristics, Y, is an inverse problem. Therefore, statistical analysis can be used to identify the best set (or sets) of Xs for a new set of observed responses Y. One can fit a model (or models) such as Υ = f(Χ) + error , for each of the responses, based on a calibration experiment and then “invert” to solve for the best set of Xs for a new set of Ys. This perspectives paper uses archived experimental data to consider aspects of data collection and experiment design for the calibration data to maximize the quality of the predicted Ys in the forward models; that is, we assume that well-estimated forward models are effective in the inverse problem. In addition, we consider how to identify a best solution for the inferred X, and evaluate the quality of the result and its robustness to a variety of initial assumptions, and different correlation structures between the responses. In addition, we also briefly review recent advances in metrology issues related to characterizing particle morphology measurements used in the response vector, Y.

36 MATERIALS SCIENCE↗

Hybrid statistical testing for nuclear material accounting data and/or process monitoring data in nuclear safeguards

The aim of nuclear safeguards is to ensure that special nuclear material is used for peaceful purposes. Historically, nuclear material accounting (NMA) has provided the quantitative basis for monitoring for nuclear material loss or diversion, and process monitoring (PM) data is collected by the operator to monitor the process. PM data typically support NMA in various ways, often by providing a basis to estimate some of the in-process nuclear material inventory. We develop options for combining PM residuals and NMA residuals (residual = measurement - prediction), using a hybrid of period-driven and data-driven hypothesis testing. The modified statistical tests can be used on time series of NMA residuals (the NMA residual is the familiar material balance), or on a combination of PM and NMA residuals. The PM residuals can be generated on a fixed time schedule or as events occur.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Estimating Alarm Thresholds for Process Monitoring Data under Different Assumptions about the Data Generating Mechanism

Process monitoring (PM) for nuclear safeguards sometimes requires estimation of thresholds corresponding to small false alarm rates. Threshold estimation dates to the 1920s with the Shewhart control chart; however, because possible new roles for PM are being evaluated in nuclear safeguards, it is timely to consider modern model selection options in the context of threshold estimation. One of the possible new PM roles involves PM residuals, where a residual is defined as residual = data − prediction. This paper reviews alarm threshold estimation, introduces model selection options, and considers a range of assumptions regarding the data-generating mechanism for PM residuals. Two PM examples from nuclear safeguards are included to motivate the need for alarm threshold estimation. The first example involves mixtures of probability distributions that arise in solution monitoring, which is a common type of PM. The second example involves periodic partial cleanout of in-process inventory, leading to challenging structure in the time series of PM residuals.

Burr, Tom↗

The Effect of Measurement Error on Surveillance Metrics

The purpose of this manuscript is to describe different simulation studies that CCS-6 has performed for the purpose of understanding the effects of measurement error on the surveillance metrics. We assume that the measured items come from a larger population of items. We denote the random variable associate with an item's value of an attribute of interest as X and that $X \text{~} N(\mu, \sigma^2$). This distribution represents the variability in the population of interest and we wish to make inference on the parameters $\mu$ and $\sigma$ or on some function of these parameters. When an item X is selected from the larger population, a measurement is made on some attribute of it. This measurement is made with error and the true value of X is not observed. The rest of this section presents simulation results for different measurement cases encountered.

42 ENGINEERING↗

Surveillance Metrics Sensitivity Study

In September of 2009, a Tri-Lab team was formed to develop a set of metrics relating to the NNSA nuclear weapon surveillance program. The purpose of the metrics was to develop a more quantitative and/or qualitative metric(s) describing the results of realized or non-realized surveillance activities on our confidence in reporting reliability and assessing the stockpile. As a part of this effort, a statistical sub-team investigated various techniques and developed a complementary set of statistical metrics that could serve as a foundation for characterizing aspects of meeting the surveillance program objectives. The metrics are a combination of tolerance limit calculations and power calculations, intending to answer level-of-confidence type questions with respect to the ability to detect certain undesirable behaviors (catastrophic defects, margin insufficiency defects, and deviations from a model). Note that the metrics are not intended to gauge product performance but instead the adequacy of surveillance. This report gives a short description of four metrics types that were explored and the results of a sensitivity study conducted to investigate their behavior for various inputs. The results of the sensitivity study can be used to set the risk parameters that specify the level of stockpile problem that the surveillance program should be addressing.

97 MATHEMATICS AND COMPUTING↗