Skip to main content
ARS Home » Research » Publications at this Location » Publication #260257
Title: Dose-Response: Background and Perspectives on the Development of Analysis Methodology

Author
item Seefeldt, Steven
item PRICE, WILLIAM - University Of Idaho
item SHAFII, BAHMAN - University Of Idaho

Submitted to: Meeting Abstract
Publication Type: Abstract Only
Publication Acceptance Date: 1/3/2010
Publication Date: N/A
Citation: N/A

Interpretive Summary: In weed science, knowledge of the amount of a chemical, fertilizer, or biological substance needed to cause a desirable response is critical to sustainable and profitable farming. There have been many problems in developing adequate analysis techniques to measure a plant response to some treatment that involves multiple quantities of a dose. Many of the statistical techniques that were developed could not be used due to limitations in computing ability. Historically, methods were developed to simplify the process using data transformation and complicated graphing methods. Advances in computer power and improvements in statistical packages have removed the computational limits. Now, more complex and biologically relevant models could now be utilized to analyze dose-response relationships.

Technical Abstract: Understanding and modelling the response of a living organism to a dose of a chemical or biological compound is an important aspect of biologic science. Many absorbed or ingested organic and inorganic components that are beneficial or at least harmless at low doses can be toxic at higher doses. When trying to recommend doses of a substance for killing a weed without harming a crop, the responses of the two plant species must be predicted. Early studies have determined that the response was often an asymmetric sigmoidal-shaped curve and that there was inherent variability in susceptibility among individuals in a population. From 1930 until almost 1960, there were heated discussions concerning whether probit or logit functions were best for modelling dose-response relationships. At that time all analyses had to be done by hand. Several attempts were made to simplify the calculations through the use of specially designed graph paper, but adoption of these methods was limited. Rather than spending large amounts of time with these models, most researchers resorted to conducting simpler average response analyses such as ANOVA at specific doses, or attempting to transform data into a linear format followed by linear regression. None of these simplified methods were useful for describing data at extreme values of doses nor were they biologically relevant. With the advent of computers and statistical software, more complex and biologically relevant models could now be utilized to analyze dose-response relationships. The estimation procedures for these models are discussed and demonstrated in the upcoming presentations in this series.