Author
 |
Seefeldt, Steven |
|
Submitted to: Western Society of Weed Science Meeting Proceedings
Publication Type: Proceedings Publication Acceptance Date: 1/5/2005 Publication Date: 2/5/2005 Citation: Seefeldt, S.S. 2005. Historical development of dose-response relationships. In: Proceedings of the Western Society of Weed Science. 58th Annual Western Society of Weed Science Meeting. March 8-10, 2005, Vancouver, Canada. CDROM. Interpretive Summary: What does a substance do to an organism? Many absorbed or ingested things (organic and inorganic) 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 known. Early studies determined that the response was often an asymmetrical sigmoidal-shaped curve and that there was an inherent variability in susceptibility among individuals in a population. From 1930 until almost 1960 there were heated discussions concerning whether probit or logistic 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 people resorted to conducting simpler statistical tests or attempting to transform data into a simpler form. None of these simplified methods were useful for describing data at high and low doses nor did they make biologic sense. With the advent of computers and statistical software, the more complex and more biologically relevant models described half a century earlier could now be utilized to analyze dose-response relationships. Technical Abstract: Understanding and modelling the response of a living organism to a dose of something is an important aspect of biologic science. Many absorbed or ingested things (organic and inorganic) 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 determined that the response was often an asymmetrical sigmoidal-shaped curve and that there was an inherent variability in susceptibility among individuals in a population. From 1930 until almost 1960 there were heated discussions concerning whether probit or logistic 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 people resorted to conducting simpler statistical tests such as ANOVA comparisons 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 extremes of doses nor were they biologically relevant. With the advent of computers and statistical software, the more complex and more biologically relevant models described half a century earlier could now be utilized to analyze dose-response relationships. |
