United States
Department of
Agriculture
Statistical
Reporting
Service
Statlstlc.1
R••••
rch
Dlvlalon
Steff R.port
Number AGES 820715
July 1982
Evaluation of the
Thompson-type
Yield Models for
Soybeans in Iowa,
Illinois, and Indiana
Richard A. Kestle
EVALUATION OF THE THOMPSON-TYPE YIELD MODELS FOR SOYBEANS IN IOWA, ILLINOIS,
AND INDIANA. By Richard A. Kestle; Research Division, Statistical Reporting
Service, U. S. Department of Agriculture, Columbia, Missouri 65201, July, 1982.
SRS Staff Report No. AGES820715
ABSTRACT
The Thompson-type models evaluated use the basic input variables of year and
monthly average temperature and total precipitation to predict soybean yields
in Iowa, Illinois, and Indiana. Both pooled and unpooled models predicting
crop reporting district and state level yields were compared; pooled models
outperformed the unpooled models.
Evaluation of yield reliability at the
state level indicated bias of pooled models less than one quintal/hectare
and standard deviation between one and two quintals/hectare.
The models are
objective.
Some input variables were not statistically significant.
Timely
yield estimates can be made with approximate (or assumed normal) climatic
division weather data. The models are not costly to use and are easy to
understand.
Key Words: Model evaluation, crop yield modeling, pooled models, regression
models.
*************************************************************
*
*
t This paper was prepared for limited distribution to the
~
: research community outside the U. S. Department of Agri-
~
: culture. The views expressed herein are not necessarily
~
: those of SRS or USDA.
:
*************************************************************
ACKNOWLEDGEMENTS
The author wishes to thank Jeanne Sebaugh, Wendell Wilson, and other AgRISTARS
Yield Model Development Project personnel for their comments and assistance~
and Jean Sparks for typing this report.
Evaluation of the Thompson-Type Yield Models
for Soybeans in Iowa, Illinois, and Indiana
Richard A. Kestle
Mathematical Statistici