Nondetects & Multivariate Statistics - 2 courses
ask...@yahoo.com
Nondetects And Data Analysis
Correctly handle data below detection limits
Nov 11-12, 2008 $795
Mainsail Suites and Conference Center, Tampa, FL
- more info at: www.practicalstats.com
For most environmental professionals, the way to deal with
"nondetects" is less than obvious. Values below detection or
quantitation limits result from measuring trace amounts of a variety
of organic and inorganic chemicals. Inadequate methods such as
substitution of one-half the detection limit, Cohen's approximate MLE,
or Aitchison's DLOG methods are still being used and recommended even
though they produce inaccurate results. This course presents up-to-
date (maximum likelihood and survival analysis) methods for computing
summary statistics, performing hypothesis tests, regression modeling
and trend analysis for data with one or more detection limits. Example
problems are worked in class, so students can confidently take these
methods back to their office. Routines in commercial software as well
as the free R statistical system are highlighted. A full course
outline is available on the Practical Stats website. This course
assumes a knowledge of basic statistics, including familiarity with t-
tests, linear regression, and simple nonparametric tests like the rank-
sum test.
INSTRUCTOR
Dr. Dennis Helsel has 30 years experience in applying statistical
methods to environmental sciences. He is author of the course textbook
Nondetects And Data Analysis (Wiley, 2005). He was awarded the
Distinguished Achievement Award by the Section on Statistics and
Environment of the American Statistical Association in 2003.
--
Untangling Multivariate Relationships
a 2-day course on applied multivariate statistics
Nov 13-14, 2008 $895
Mainsail Suites and Conference Center, Tampa FL
- more info at: www.practicalstats.com
Environmental measurements often are simultaneously recorded as a
series of variables or species. Multiple chemicals are analyzed.
Multiple measures of ecosystem health are recorded. If statistics are
computed on each, one by one, the interrelations among variables,
species and locations cannot be seen. Valuable information is lost.
Multivariate methods resolve what at first may look like noise into
recognizable patterns, providing new insight into the field of study.
However, these methods are daunting to many scientists, with acronyms
like PCA, CCA or CANOCO, and with unfamiliar terms like varimax
rotation and detrended correspondence. The learning curve seems steep,
and useful procedures go unused.
UMR covers the multivariate methods of primary interest to
environmental science, focusing on what each method is designed to do,
when to use them, and when not to. Methods for simplifying and
plotting data (ordination) are contrasted with those for establishing
connections between and among source and response variables.
Capabilities of various software packages are reviewed. Example data
sets are analyzed by each student in class. Our goal is that students
leave with a clear understanding of which method to use, and how to
use it, for their own data.
--