Relationships between weather and downy mildew epidemics of hop in the
Yakima Valley of Washington State over a 38-yr period were examined w
ith linear discriminant and logistic regression analyses. A linear dis
criminant function with minimum temperature in April X the total numbe
r of rainfall days in April and May, an indicator variable for the pre
sence of an epidemic the preceding year, and the minimum temperature i
n May X the number of occurrences of three or more consecutive rainfal
l days in May correctly classified the epidemic status for 92% of the
years. The percentage of years with mild epidemics correctly classifie
d (specificity) and years with severe epidemics correctly classified (
sensitivity) for a 10-yr test data set were 86 and 10 %, respectively.
Logistic regression, which requires fewer assumptions, had results ve
ry similar to those of the discriminant analysis.