Robert A. Lordo

An enzyme-linked immunosorbent assay for the determination of dioxins in contaminated sediment and soil samples.

A 96-microwell enzyme-linked immunosorbent assay (ELISA) method was evaluated to determine PCDDs/PCDFs in sediment and soil samples from an EPA Superfund site. Samples were prepared and analyzed by both the ELISA and a gas chromatography/high resolution mass spectrometry (GC/HRMS) method. Comparable method precision, accuracy, and detection level (8 ng kg(-1)) were achieved by the ELISA method with respect to GC/HRMS. However, the extraction and cleanup method developed for the ELISA requires refinement for the soil type that yielded a waxy residue after sample processing. Four types of statistical analyses (Pearson correlation coefficient, paired t-test, nonparametric tests, and McNemars test of association) were performed to determine whether the two methods produced statistically different results. The log-transformed ELISA-derived 2,3,7,8-tetrachlorodibenzo-p-dioxin values and log-transformed GC/HRMS-derived TEQ values were significantly correlated (r=0.79) at the 0.05 level. The median difference in values between ELISA and GC/HRMS was not significant at the 0.05 level. Low false negative and false positive rates (<10%) were observed for the ELISA when compared to the GC/HRMS at 1,000 ng TEQ kg(-1). The findings suggest that immunochemical technology could be a complementary monitoring tool for determining concentrations at the 1,000 ng TEQ kg(-1) action level for contaminated sediment and soil. The ELISA could also be used in an analytical triage approach to screen and rank samples prior to instrumental analysis.

Assessment of relative potential for Legionella species or surrogates inhalation exposure from common water uses

The Legionella species have been identified as important waterborne pathogens in terms of disease morbidity and mortality. Microbial exposure assessment is a tool that can be utilized to assess the potential of Legionella species inhalation exposure from common water uses. The screening-level exposure assessment presented in this paper developed emission factors to model aerosolization, quantitatively assessed inhalation exposures of aerosolized Legionella species or Legionella species surrogates while evaluating two generalized levels of assumed water concentrations, and developed a relative ranking of six common in-home uses of water for potential Legionella species inhalation exposure. Considerable variability in the calculated exposure dose was identified between the six identified exposure pathways, with the doses differing by over five orders of magnitude in each of the evaluated exposure scenarios. The assessment of exposure pathways that have been epidemiologically associated with legionellosis transmission (ultrasonic and cool mist humidifiers) produced higher estimated inhalation exposure doses than pathways where epidemiological evidence of transmission has been less strong (faucet and shower) or absent (toilets and therapy pool). With consideration of the large uncertainties inherent in the exposure assessment process used, a relative ranking of exposure pathways from highest to lowest exposure doses was produced using culture-based measurement data and the assumption of constant water concentration across exposure pathways. In this ranking, the ultrasonic and cool mist humidifier exposure pathways were estimated to produce the highest exposure doses, followed by the shower and faucet exposure pathways, and then the toilet and therapy pool exposure pathways.

Evaluating the effectiveness of state specific lead-based paint hazard risk reduction laws in preventing recurring incidences of lead poisoning in children☆

BACKGROUND Despite significant progress made in recent decades in preventing childhood lead poisoning in the United States through the control or elimination of lead sources in the environment, it continues to be an issue in many communities, primarily in low-income communities with a large percentage of deteriorating housing built before the elimination of lead in residential paint. The purpose of this study is to determine whether state laws aimed at preventing childhood lead poisoning are also effective in preventing recurring lead poisoning among children previously poisoned. METHODS An evaluation was conducted to determine whether laws in two representative states, Massachusetts and Ohio, have been effective in preventing recurrent lead poisoning among children less than 72 months of age previously poisoned, compared to a representative state (Mississippi) which at the time of the study had yet to develop legislation to prevent childhood lead poisoning. RESULTS Compared to no legislation, unadjusted estimates showed children less than 72 months old, living in Massachusetts, previously identified as being lead poisoned, were 73% less likely to develop recurrent lead poisoning. However, this statistically significant association did not remain after controlling for other confounding variables. We did not find such a significant association when analyzing data from Ohio. CONCLUSIONS While findings from unadjusted estimates indicated that state lead laws such as those in Massachusetts may be effective at preventing recurrent lead poisoning among young children, small numbers may have attenuated the power to obtain statistical significance during multivariate analysis. Our findings did not provide evidence that state lead laws, such as those in Ohio, were effective in preventing recurrent lead poisoning among young children. Further studies may be needed to confirm these findings.

A Screening Analysis for Incomplete Ranking Data

We consider the setting where a group of n judges are to rank independently a series of k objects. When all judges rank all treatments, this coincides with the no-interaction two-way lay-out setting with one observation per cell. However, many times the intended complete rankings are not realized and we are faced with analysing randomly incomplete rank vectors. We propose a new screening approach to analysing such data realizations. For discussion, we concentrate on the problem of testing for no differences in the objects being ranked (i.e. they are indistinguishable) against general alternatives, but our approach will also be quite appropriate for restricted (e.g. ordered or umbrella) alternatives. We detail the algorithm for computation of the relevant Friedman-type statistic in the general alternatives setting and present the results of a simulation study to assess the effectiveness of the algorithm.

Nonparametric and Semiparametric Models

Weisberg (1985). Also, very little recent literature (after 1984) is covered (with the exception of Sec. 7.3, which covers radial basis functions). Following Cook and Weisberg (1999, p. 432), the most important idea from the recent literature is that MLR is the study of the conditional distribution of the response variable given the predictors, and this distribution can be visualized with a plot of the fitted values versus the response variable. Texts that do not discuss this plot may be obsolete.

Image Processing and Jump Regression Analysis

My dictionary defines algorithm broadly as “a step-by-step procedure for solving a problem or accomplishing some end” and thus supports what I thought was a misuse of the term in the subtitle of this book. The authors, both faculty members at the University of Waterloo, got a lot more than that right in this extraordinary book devoted to the systematic reduction in variability. The book is an outgrowth of the authors’ work as participants in the University’s Institute for Improvement in Quality and Productivity (IIQP). They acknowledge the support and participation of nearly 30 companies as diverse as GM Canada and Campbell soup. I learned from the website that IIQP has since been renamed The Business and Industrial Statistics Research Group and that Professor Steiner is the Director. The book is used as the text in a course in Productivity Improvement at the University and comes with a CD containing 109 datasets in both Excel and Minitab formats that came from actual variance reduction projects described in the text. The CD also contains 356 pages of additional material including exercises and their solution, supplementary material for 11 of the book’s 21 chapter and six appendixes devoted to the use of Minitab. The book is a bonanza for those searching for relevant examples for a short course for a manufacturing staff or to enliven a course in Engineering Statistics. The book is targeted at chronic, as opposed to sporadic, problems of excessive variability met in manufacturing and assembly, in medium to high volume production environments. The implementation of the algorithm is presumed to be delegated to a team whose members possess process and statistical knowhow. The authors enumerate seven ways, of varying complexity, in which process variability may be reduced. The algorithm itself is set forth in flow chart format. The first step is to define a focused problem. The next step is to check the adequacy of the measurement system. Following this step the team proceeds to select one of the seven ways to reduce variation. These divide into three ways requiring the identification of a dominant cause and four ways which do not. The authors’ experience teaches that in most cases the team will decide to search for a dominant cause. The algorithm loops until a choice of variation reduction method is shown to be feasible whereupon a plan for implementation is developed. The last step comprises three elements: (i) implementing the solution, (ii) validating the solution, and (iii) holding the gains, that is, putting monitoring techniques in place so the reduction in variability is maintained over time. In Chapter 1, “Introduction,” the authors describe seven real world problems such as a case of excessive variability in a crankshaft diameter and a problem with engine block leaks. They present what they call the baseline data, typically histograms or scatter plots and summary statistics that define the situation as the team found it. These same problems are frequently referred to in subsequent chapters as the algorithm is applied to them. Chapters 2–5 constitute Part I, entitled “Setting the Stage.” Chapter 2 is called “Describing Processes.” Chapter 3, called “Seven Approaches to Variation Reduction,” describes and illustrates, with examples, each of the seven ways to reduce variability. Chapter 4 describes the algorithm itself. Chapter 5, “Obtaining Process Knowledge Empirically,” gives a framework for learning about a process, to which the authors apply the acronym QPDAC, standing for Question, Plan, Data, Analysis, and Conclusion. Two examples illustrate the QPDAC procedure, one involving a piston diameter measurement system and the other involving an unwanted residual pattern in a two-color laminating process. Chapters 6–8 amplify respectively on the first three steps of the algorithm and comprise Part II of the book captioned “Getting Started.” Chapter 6 addresses the conduct of the baseline investigation and contains some straight talk on sample size requirements. The authors advocate samples large enough that sampling error is negligible which means “hundreds of units if the output characteristic is continuous and thousands if the output is binary.” They also suggest for this reason, that binary variables be converted to continuous where possible substituting for example, the measured thickness of an engine wall for the binary “leaks or doesn’t.” Part III, called “Finding a Dominant Cause of Variation,” comprises Chapters 9–13. The idea of families of causes of variation are considered such as variation that occurs over time and variation that occurs either before or after a certain point in the process. A deductive elimination method worthy of Sherlock Holmes is used to nail the culprit. Statistical techniques appearing in these chapters include histograms, scatter plots, box plots, regression and, in Chapter 13, a 23 factorial experiment, with a Pareto chart to assess the relative magnitude of effects. Statistical orthodoxy is subservient to practical results so there is no obsession with using random samples, hypothesis tests, confidence limits or tests of normality. Process knowledge and a few statistical tools wisely applied are combined in the common goal of finding and eliminating variation. Part IV, entitled “Assessing Feasibility and Implementing a Variation Reduction Approach,” comprises Chapters 14–21. Moving a process center by finding an input that acts as an “adjuster” is dealt with as is desensitizing a process to variation and its close relative, making a process robust. Feedforward control, feedback control, and 100% inspection round out the options. One refreshing thing about the book is that not every project described was a success. Usually every project described has valuable lessons for ultimate success. In one case a team ran an experiment to find the combination of six variables that would result in the smallest variability. When implemented there was no improvement because the short term variability seen in the experiment was a small component of the longer term variability shown in the baseline data. In other cases the only possible cure was too expensive to implement and the project was abandoned. In summary, this book offers systematic guidance for practitioners and practical illustrative material for teachers. It had its genesis in a most salutary collaboration of those two groups.

Statistical Reasoning and Methods

About Statistics. Organizing Data and Describing Patterns. Describing Bivariate Data. ProbabilityThe Basis for Inference. Random Variables and Probability Distributions. Normal Distributions. Variation in Repeated SamplesSampling Distributions. Inferences About MeansLarge Samples. Small Samples Inferences for Normal Populations. Comparing Two Treatments. Analyzing Count Data. Regression AnalysisSimple Linear Regression. Appendices. Data Bank. Answers to Selected Odd-Numbered Exercises. Index.