Multivariate statistical analysis of environmental monitoring data

Journal Article · Sat Nov 01 04:00:00 EST 1997 · Ground Water

DOI:https://doi.org/10.1111/j.1745-6584.1997.tb00177.x· OSTI ID:556599

Ross, D L ^[1]

---

1. Glenrose Engineering, Austin, TX (United States)

EPA requires statistical procedures to determine whether soil or ground water adjacent to or below waste units is contaminated. These statistical procedures are often based on comparisons between two sets of data: one representing background conditions, and one representing site conditions. Since statistical requirements were originally promulgated in the 1980s, EPA has made several improvements and modifications. There are, however, problems which remain. One problem is that the regulations do not require a minimum probability that contaminated sites will be correctly identified. Another problems is that the effect of testing several correlated constituents on the probable outcome of the statistical tests has not been quantified. Results from computer simulations to determine power functions for realistic monitoring situations are presented here. Power functions for two different statistical procedures: the Student`s t-test, and the multivariate Hotelling`s T{sup 2} test, are compared. The comparisons indicate that the multivariate test is often more powerful when the tests are applied with significance levels to control the probability of falsely identifying clean sites as contaminated. This program could also be used to verify that statistical procedures achieve some minimum power standard at a regulated waste unit.

• Sponsoring Organization: USDOE
• OSTI ID: 556599
• Journal Information: Ground Water, Journal Name: Ground Water Journal Issue: 6 Vol. 35; ISSN GRWAAP; ISSN 0017-467X
• Country of Publication: United States
• Language: English

Related Subjects

54 ENVIRONMENTAL SCIENCES
GROUND WATER
MONITORING
MULTIVARIATE ANALYSIS
SOILS
WASTE WATER
WATER POLLUTION