ST01 Analysis of Dosage-Response Data in Agricultural Research     Contributed

Dr. Khorsand Bondari
University of Georgia
Abstract: Agricultural researchers, especially entomologists, are interested in determining how mortality rates change with increasing dose levels of a certain stimulus (e.g., pesticide, drug). A researcher may have a particular interest in determining the pesticide dose at which 50% (LD50), 75% (LD75), or 90% (LD90) of insect populations respond. Several procedures have been suggested for estimating the best dosage choice for an insecticide. When the response variable is binary or measured ordinally rather than cont inuously, PROBIT or LOGIT analyses based on Maximum Likelihood procedure are appropriate. This report will explore the use of several dose-response models including LOGISTIC, GENMOD, PROBIT, and CATMOD procedures of SAS/STAT which can all be used for stati stical modeling of the dose-response data. All these methods predict the probability of a positive response (death) as a function of the pesticide dosage applied. Similar approaches have been widely used in many fields of research including medicine, labor atory animals research, economics, sociology, and genetics.

Biography:
Dr. Bondari obtained his M.S. and Ph.D. from Iowa State University in Ames, IA. He has been a University of Georgia faculty for 25 years and has a joint faculty appointment in the Department of Entomology (20%) and Experimental Statistics Unit (80%) at the rank of Professor. He serves as a statistical consultant for agricultural research and extension scientists at the University of Georgia’s Tifton-Campus at the Coastal Plain Experiment Station. He has organized and chaired sessions at the 1992 and 2001 (S tatistical Consulting Section) and 1995 (Statistics and the Environment Section) Joint Statistical Meetings and has served as a coordinator in SUGI 25, SUGI 26, and SUGI 27 for the Statistics, Data Analysis, and Data Mining Section.