Practical Stats Training

A free online course in environmental statistics, at your desk and on your time using real-world data.

The online course Nondetects And Data Analysis (NADA) teaches statistical methods for computing summary statistics, hypothesis tests, regression, and more, all for data with nondetects while avoiding the use of fabricated values such as 1/2 the detection limit. Adding fabricated data causes many (usually unintended) problems such as incorrect estimates of the mean, incorrect hypothesis test results, and sub-optimal regression equations. Instead, methods designed for censored data (data with nondetects in this case) avoid these problems and produce far more reliable and defensible results. After great expense for salaries, field collection and laboratory analysis of low-level contaminants, do you really want to make up values for part of your data and then try to extract the information? To understand the problems that occur when fabrication of various sorts is used, see below either the short video of Why Not Substitute Values for Nondetects? or the longer webinar The Cost of Complacency.

There are better ways, and you'll find them demonstrated on real-life field datasets in our NADA course. Hundreds of articles have used the procedures in NADA and NADA2 to evaluate their data without substituting fabricated values for nondetects. A list of only a few of these articles are on our NADA2 references page.