Ms. Edwards is an accredited statistician and has testified on the use of statistical methods. She regularly performs and critiques data analytics of lab chemistry concentrations, toxicity tests, species community structure, field screening results, and background or reference comparisons. She is called upon frequently to provide understandable descriptions of statistical methods used by testifying experts and to assist with deposition questioning. Her areas of application are expansive, including chemical forensics, pesticide registration, and injury quantification. With over 20 years of experience, Ms. Edwards has provided statistical support on projects involving metals, PCBs, PAHs, and dioxins/furans in soil, sediment, dust, groundwater, surface water, and tissue.
She routinely provides insight on data presentation and interpretation of statistical analyses for experts from a variety of backgrounds, including environmental science, ecology, toxicology, and engineering as well as non-scientists. She is also familiar with public data sources such as NHANES and SEER data repositories as well as chemistry datasets representing background concentrations in soil and water. She has maintained all aspects of databases used for data validation, compilation, and storage, including transfer of data from multiple formats. Ms. Edwards has provided statistical support on a wide range of projects including chemical fingerprint analyses, model fitting and predictions, comparisons with background and reference populations, probabilistic model development for quantification of uncertainties, evaluation of toxicity test results and sampling program methodologies Traditional methods she uses include regression, analysis of variance, non-linear models, factor analysis, principal component analysis, non-parametric methods, and sampling design and evaluation. Additionally, she is familiar with modern data analytics techniques as they apply to environmental data.
CREDENTIALS & PROFESSIONAL HONORS
- M.S., Statistics, University of Wisconsin, Madison, 1998
- B.S., Mathematical Science, University of Washington, 1991
LICENSES & CERTIFICATIONS
Accredited Professional Statistician, PStat®, American Statistical Association
O’Reilly KT, Edwards M. Letter to the Editor: Comment on Norris and Henry (2019). Science of the Total Environment 2019.
Edwards M, Pietari J, Cook L, Boehm P. Treatment of non-detects can lead to inaccurate forensic conclusions. Detection Limits in Air Quality and Environmental Measurements, ed. M.J. Brisson. ASTM International 2019; 181-191.
Morrison AM, Edwards M, Buonagurio J, Cook L, Murray K, Boehm P. Assessing the representativeness and sufficiency of water samples collected during an oil spill. Proceedings, 2017 International Oil Spill Conference, Vol 2017, No 1.
Shields W, Edwards M, Monti X, Royer L. Examination of PCDD/F outliers in soil background studies provides insights into historic sources. Organohalogen Compounds 2016; 78:576-578
Aldea M, Edwards M, Pietari J, Boehm P. Cautions on the treatment of non-detect results for environmental forensics. Environmental Forensics 2016; 17(4):311-318.
Menzie C, Deardorff TL, Ma J, Edwards M. Risk factors that contribute to the occurrence of catastrophic wildfires in California. World Environmental and Water Resources Congress 2015; 2617–2626.
Monti C, Shields W, Edwards M, Pozzi C. Fingerprint analysis of PCDD/F in soil samples in the Taranto (Puglia, Italy) area. Organohalogen Compounds 2014; 76:830–833.
Aylward L, Monti C, Edwards M, Shields W. Dioxins in sheep and goat liver tissues from Taranto, Italy: Congener pattern. Organohalogen Compounds 2014; 76:403–406.
Shields W, Edwards M. Comparison of dioxin/furan characteristics in different US cities. Abstract 426–2, Water, Food, Energy & Innovation for a Sustainable World, ASA, CSSA, & SSA International Annual Meetings, Tampa, FL, November 3–6, 2013. Available at https://scisoc.confex.com/crops/2013am/webprogram/Paper81402.html.
Megson D, O’Sullivan G, Comber S, Worsfold PJ, Lohan MC, Edwards MR, Shields WJ, Sandau CD, Patterson DG Jr. Elucidating the structural properties that influence the persistence of PCBs in humans using the National Health and Nutrition Examination Survey (NHANES) dataset. Science of the Total Environment 2013; (461–462):99–107.
Moolgavkar SH, McClellan RO, Dewanji A, Turim J, Luebeck EG, Edwards MR. Time-series analyses of air pollution and mortality in the United States: a subsampling approach. Environmental Health Perspectives 2013; 121(1):73–78.
Shields WJ, Edwards MR, Abrahams JA, Ferrara RA, Bollinger M, Paul LS. Source evaluation of dioxins/furans, PAHs, and pentachlorophenol in soil samples near a wood treating site in the Southeastern USA. Organohalogen Compounds 2012; 74:593–595.
Connor KT, Harris MA, Edwards MR, Budinsky RA, Clark GC, Chu AC, Finley BL, Rowlands JC. AH receptor agonist activity in human blood measured with a cell-based bioassay: Evidence for naturally occurring AH receptor ligands in vivo. Journal of Exposure Science and Environmental Epidemiology 2008; 18:369–380.
Lowney YW, Wester RC, Schoof RA, Cushing CA, Edwards MR, Ruby MV. Percutaneous absorption of arsenic from soils as measured in the Rhesus Monkey. Toxicological Sciences 2007; 100(2):381–392.
Benton, L, Shields, WJ, Edwards, MR. Commentary on O'Connor and Sabrsula (2005): Background dioxins in house dusts. Environmental Forensics 2007; 6(3):283–287.
Tsuji JS, Williams PRD, Edwards MR, Allamneni KP, Kelsh MA, Paustenbach DJ, Sheehan PJ. Evaluation of mercury in urine as an indicator of exposure to low levels of mercury vapor. Environ Health Perspect 2003; 111(4):623-630.
Shields WJ, Tondeur Y, Benton L, Edwards MR. Dioxins and furans. Chapter 14. In: Environmental Forensics: Contaminant Specific Guide. Morrison R, Murphy B (eds), Academic Press, San Diego, CA, 2006.
Edwards, M., Pietari, J. Treatment of non-detects in environmental forensics: incorrect conclusions?! ASTM Symposium on Detection Limits, Washington, D.C., October 25-26, 2018.
Edwards, M. Sparacio, T., Shields, W. Successful use of visual forensics at a smelter site. INEF conference, Salt Lake City, UT, June 25-27, 2018.
Edwards, M, Aldea M. Effect of data transformations on the interpretability of principal component analysis results. Poster presentation, Joint Statistical Meeting, Seattle, WA, August 8–13, 2015.
Aldea, M, Edwards M, Pietari J. Cautions on the treatment of non-detect results for environmental forensics. Poster presentation, INEF conference, Toronto, Canada, August 3–6, 2015.
Aldea M, Edwards M, Buonagurio J. Visualizing the potential of environmental big data. Presentation at the 8th International Conference on Remediation and Management of Contaminated Sediments, New Orleans, LA, January 2015.Edwards M. The significance of statistical significance. Presentation at the DRI Annual Meeting, New Orleans, LA, 2012.
Shields WJ, Edwards, MR, Abrahams JA, Ferrara, RA, Bollinger M, Paul LS. Source evaluation of dioxins/furans, PAHs, and pentachlorophenol in soil samples near a wood treating site in the Southeastern USA. Dioxin 2012, 32nd International Symposium on Halogenated Persistent Organic Pollutants, Cairns, Queensland, Australia, 2012.
Menzie C, Deardorff T, Ma J, Edwards M. California’s catastrophic wildland fires: Increasing the risks of burning hotter, faster, and higher. Poster presentation, SETAC North America 33rd Annual Meeting, Long Beach, CA, November 11–15, 2012.
Johns M, Atlas R, Edwards M. Weathering of MC252 oil in the water column of the Gulf of Mexico from May through September 2010 : Fluorometry, dissolved oxygen, and quantitative chemistry evaluation. Poster presentation at the Gulf Oil Spill SETAC focused topic meeting, Pensacola, FL, 2011.
Edwards M, Maier E, Reeder D, Gard N, Shock S. Assessment of plant communities exposed to fugitive dust along a mine transportation corridor in Alaska. Poster presentation at the Annual Joint Statistical Meeting, Salt Lake City, UT, 2007.
Benton L, Ginevan M, Edwards M. Chemical fingerprinting of dioxins using a similarity index. Poster presentation at the 16th Annual AEHS Meeting and West Cost Conference on Soils, Sediments and Water, San Diego, CA, 2006.
Kaetzel R., Edwards M, Lowney Y, Tsuji J. Patterns of urinary excretion of arsenic metabolites in a population with low arsenic exposure. Poster presentation at the 45th Annual SOT Meeting, San Diego, CA, 2006.
Ginevan M, Edwards M, Benton L. Statistical tools for the analysis of PCB and dioxin profile data. Platform presentation at the SETAC 26th Annual Meeting in North America, Baltimore, MD, 2005.
Pastorok R, Davis R, Salatas J, Edwards M. Identifying structural and functional indicators for habitat equivalency analysis (HEA) and restoration of submerged vegetation. Interactive platform presentation at the SETAC World Congress and 25th Annual Meeting in North America, Portland, OR, 2004.
Booth, P, Edwards M, Nielsen D. Spatial and temporal trends in PCBs in fish in response to natural recovery in sediments. Poster presentation at the SETAC World Congress and 25th Annual Meeting in North America, Portland, OR, 2004.
American Statistical Association
Provided evidence of localized source influences through principal component analysis of dioxin and furan concentrations in soil near a former chemical manufacturing plant in West Virginia. Results demonstrated no broad impacts to the surrounding residential area. Further chemical forensic analyses based on blood levels of residents showed dioxin/furan compositions of former plant workers distinct from other residents.
Used principal component analysis of dioxin and furan concentrations from multiple locations and sources to determine if historic releases from a wood treatment facility were the source of these compounds found in a nearby neighborhood in Mississippi. Results were included in the expert report of a top soil scientist.
Compared diagnostic ratios to evaluate sources of polycyclic aromatic hydrocarbons (PAH). Analyses provided the basis for a court-submitted expert report.
Developed discriminant analysis to predict likely source contributions based on selected parameters characterizing groundwater compositions in Nevada. Results and sensitivity assessment were presented at an executive meeting of responsible parties because conclusions contradicted prior understanding presented by other responsible parties.
Used profile comparisons and principal component analysis to evaluate dioxin and furan fingerprints in water, sediment, and fish tissue samples downstream and upstream of a former pesticide manufacturing facility. Correlation and cosine-theta methods quantified the similarity between composition profiles utilizing multiple standardizations.Analyzed mercury concentrations in homes using discriminant analysis to characterize and differentiate between sources of mercury, including natural gas pressure regulators, thermometers, and other residential sources of mercury.
Exposure and Effect Modeling
Random forest analysis provided quantitative support that no combination of the multitude of chemicals measured in sediment near a metals smelter site were associated with toxic responses (survival, biomass, and growth) of Hyalella and Chironomid species.
Complex dose-response relationships of cholinesterase inhibition across multiple studies to quantify benchmark dose and uncertainty safety limit. Model results supported sensitivity assessment between adults and pups and across gender.
Generated upper tolerance limits (UTL) for concentrations of PCB congeners, dioxin and furan congeners, and toxicity equivalents (TEQ) based on the NHANES complex survey study design. These values represent the upper limit of blood concentrations for background exposure levels in the U.S. population. Comparisons provided evidence to conclude limited to no excess exposure at residential areas near a chemical manufacturing site and wood treatment facility.
Used a generalized linear model (with Poisson link function) and regression analyses to model dose-response relationships for mink mortality and reproduction (decline in fecundity). Calculated cumulative PCB dose probability curves and exceedance probabilities for specified response limits.
Calculated multiple benthic metrics (abundance, richness, diversity, and dominance) and assessed statistically significant differences between site and reference areas to support a weight-of-evidence approach to assessing biological effects, and support calculation of apparent effects threshold values.
Analyzed chemical endpoint data from surrogate species for human and wildlife receptors to quantify the relative uptake of metals from environmental media, following exposure via dermal or oral contact. Final analyses quantified the relative bioavailability of various treatments against a control. Additionally, the sample design was evaluated with respect to the minimum detectable uptake.
Used logistic regression to characterize the relationship between PAH concentration and average mortality of Hyalella Azteca.
Relationships between Environmental Media and Chemicals
Used random forest to support measurement of additional parameters in groundwater near the waste ponds of a coal power plant. Analysis quantified the limit of available data to explain variability in the distribution of arsenic concentrations.
Developed data assessment and quantification analysis to determine whether field collected fluorescence and dissolved oxygen measurements in the water column were associated with lab quantified benzene, ethylbenzene, toluene, xylene (BTEX) and PAH concentrations. Complications overcome included multiple instruments and data processing techniques. Designed the database and provided oversight for data compilation from multiple diverse sources encompassing three million sample records each with multiple measurements.
Compiled data from multiple studies of mercury in air and urine to assess common relationships using regression analysis. Adjustments of slopes and intercepts allowed for assessing similarity of sample design among studies. Confidence and prediction intervals quantified uncertainty over the range of the fitted relationships.
Developed a compilation metric incorporating vegetation, metals enrichment, physical, and geologic characteristics to identify alternative reference areas that characterize conditions prior to mining activities in northern Alaska. The derived metric quantified the similarity of conditions or lack thereof with historical reference areas used by other researchers.
Evaluated the relationship between PCB congener concentrations and total PCB concentrations in sediment and several fish species from the Fox River, Wisconsin.
Used regression analysis to evaluate the results of two analysis methods used to measure mercury concentration in sediment and soil. Results provided statistical support for using both methods, thereby increasing the quantity and precision of data used for site characterization.
Spatial and Temporal Trends
Evaluated and interpreted spatial and temporal trends in sediment PCB composition and concentration throughout a 25-mile stretch of river, in support of a NRDA for a Superfund site in Wisconsin.
Quantified temporal trends in concentrations of multiple parameters measured in groundwater monitoring wells surrounding historical holding ponds used by an electric power generating facility. Evaluated whether contributions to surface water were detectable from specific portions of the site.
Quantified the lack of a consistent spatial pattern expected from aerial deposition sources using correlation and regression methods. Analyses included distance from the potential source in question as well as alternative sources. Similar approaches have been used for sources of lead, arsenic, and dioxin and furan congeners.
Evaluated temporal trends in landfill groundwater monitoring concentrations using summary statistics, boxplots, probability plots and distance weighted least squares regression. Standardized concentrations were used to equate the varying magnitude and variation of multiple chemicals.
Programming, Data Management, and Modeling
Manager and technical reviewer for litigation support related to health effects from limitation of daylight exposure. Linked American Time Use Survey (ATUS) with location-specific astronomical data to estimate average daylight exposure times. Compared survey sample demographics with census statistics for the urban site area to assure estimates were representative.
Developed statistical analysis code in R to calculate percentile estimates of reference levels of exposure representative of the general US population based on the complex probabilistic sampling design underlying the NHANES publicly available data.
Developed probabilistic model to manage global waste stream volumes for a large petroleum company, including liquid and solid waste from sites ranging in size from local gas stations to major port loading facilities. Results from the model provided a baseline volume estimate prior to implementation of waste reduction practices and more detailed record keeping.
Managed large databases of chemistry, toxicity, and ecological data for multiple projects throughout the US and extracted compatible subsets of data for statistical analysis.
Study Design, Minimum Detectable Change, and Statistical Power
Provided guidance to top scientific researcher for appropriate statistical analysis to assess differences in treatment effects of reducing mercury methylation in sediment. Prepared figures and summary tables for the final client report.
Used simulation methods to evaluate sources of variability in fish tissue PCB and mercury concentrations and evaluate the state of Michigan’s fish monitoring program. Recommended changes to the sample design that would achieve the equivalent monitoring information at lower costs.
Determined dermal absorption of arsenic from soil are not different from background levels using analysis of variance (ANOVA) with nested factors. Additionally, evaluated statistical power of study design and quantified minimum detectable absorption as recommendation to U.S. EPA for updating the human health risk assessment guidance for arsenic.
Evaluated water quality monitoring program design, implementation, and data for ability to characterize runoff from agricultural lands and suitability in triggering implementation of management practices to minimize water quality impacts of agricultural practices.
Calculated the minimum number of samples necessary to meet statistical performance objectives of sampling plan in accordance with EPA guidance, including multiple levels of confidence, power, and minimum detectable differences.
News & Events