This workshop will explain the fundamental mismatch between common (classical) statistics and its assumptions vs. the nature of chemically contaminated environmental media. For example, common statistical approaches assume that contamination is randomly and uniformly distributed at the spatial scales typical for sample collection and chemical analysis. Data from case examples will show this is not the case. When the effects of heterogeneity are not controlled, misleading data can be the result with the possibility of leading to faulty decision-making. Other limitations of classical statistics for determining sample numbers and locations will be reviewed, along with consideration of common questions, such as “What is the ‘gray region’ and how do I set it?” and “How do I know if a sample is ‘representative’?”
The importance of conceptual site models (CSMs) and “sample support” for effective sampling design will be emphasized. A good understanding of the CSM and the specifics for how regulatory thresholds will be applied is required BEFORE a sampling plan can be designed. Real-time data generation is an important strategy since it supports gathering higher densities of data from the most informative locations, thereby reducing the likelihood of non-representative data. Armed with this information, the participant will be able to decide when classical statistics is appropriate for the task at hand, and how to adapt classical statistical designs to improve their performance. Finally, the course will provide an overview of sampling design alternatives that are better suited to deal with the difficulties posed when generating data from real-world matrices.
A CD-ROM with the PPT slides and various resources will be provided to participants.
| Upcoming Events
No upcoming events.
Previous Events (click to view/hide)