Search NASA⌕ Search

Engineering topics

Velsko, Stephan P.

Publications and source records attributed to Velsko, Stephan P..

Using Dempster-Shafer theory to assess explanations for the Zimbabwe anthrax outbreak of 1978-79

I use a simple version of Dempster-Shafer theory (DST) to evaluate evidence from the anthrax outbreak in Zimbabwe (Rhodesia) during 1978-9 and assess the question of whether the outbreak arose naturally or was caused deliberately. The method is derived from a DST-based “opinion calculus” developed by Josang and others in which each item of evidence is used to elicit an opinion about the focal hypothesis, in this case the assertion that the outbreak was deliberately generated. I use a verbal probability expression mapping method to elicit opinions as vectors expressing degrees of belief, disbelief and uncertainty. Two rules for combining the evidence to produce an overall opinion are compared. Both rules result in an opinion favoring the hypothesis that the outbreak was natural in origin. Many items of evidence that have been proffered over the years are shown to be nearly irrelevant to the final conclusion, while the absence of certain expected types of evidence plays a critical role in the assessment.

59 BASIC BIOLOGICAL SCIENCES↗

Conditional Dependence in Microbial Forensic Assays - A Primer

This report provides an introduction to the topic of conditional dependence in the context of microbial forensic assays. Conditional dependence between two items of evidence E1 and E2 occurs when they are both used to support a hypothesis, but E1 affects the probability of E2 and vice versa. Ignoring this dependence can lead to very large errors in estimating the diagnosticity of the combined evidence. To introduce readers to this concept, a number of definitions of conditional dependence that have been used by authors in the past have been collected together and compared. Formal mathematical relationships that constrain conditional dependence are summarized. There are several specific scenarios in which unrecognized conditional dependence can arise in microbial forensic contexts. This report provides some notional examples that illustrate dramatic effects of conditional dependence on the weight of microbial forensic evidence, and discusses the relevance of these observations for the validation of microbial forensic assays. A two-­parameter model that describes the transition between various limiting forms of conditional dependence relations is provided in an appendix.

59 BASIC BIOLOGICAL SCIENCES↗

Validation Strategies for Microbial Forensic Analysis

The validation of the probative value of microbial forensic techniques is a critical aspect of research and development that requires careful planning and a sound conceptual framework. This paper outlines a particular approach to the validation of certain types of forensic methods that naturally generates statistical measures of the relevance and weight of the scientific evidence derived from them. The suggested approach is based on the likelihood ratio interpretation of Federal Rules of Evidence 401 and 402 and allows measurement evidence to be to be presented in a format that resembles the forensic "gold standard" - human DNA typing. Examples of specific genetic and chemical and physical analysis methods are used to illustrate how this general strategy can be applied. This approach also provides a natural interpretation of the notion of "preliminary validation" that has been proposed in the literature.

36 MATERIALS SCIENCE↗

Draft SOP for Sample Matching using ICP-MS

The purpose of this standard operating procedure (SOP) is to provide an assay for testing the hypothesis that two samples are from the same sub-population, which can mean they came from the same batch, set of replicate batches, or the same process. The outcome is a likelihood ratio, which should be used in combination with prior odds to determine the posterior odds of the two samples coming from the same sub-population. Note that the term “match” is not used because it is imprecise in the context of determining the relationship between two samples. Under this SOP, inductively coupled plasma mass spectrometer (ICP-MS) is used to generate elemental composition data, which is used to test the hypothesis that two samples come from the same sub-population. ICP-MS is a highly sensitive bulk analysis method. The quantity of sample required is ultimately determined by the instrumentation used and the elements selected for generating ROC curves. This SOP specifies that greater than 5 mg of sample be used to enable accurate weights to be obtained for the sample in tared 3 mL Teflon vials. However, we have conducted preliminary assays with our instrumentation that suggest that sample quantity can be reduced to 50 micrograms if elemental abundances are normalized to total metal content (see Step 10). We use ICP-MS because it is the most sensitive method for determining the concentration of the largest number of elements. Other elemental analysis methods could be used to generate data for this test (e.g., ICP-OES), but specific tests would be necessary to work out the SOP.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗