Natural variation in baseline data: when do we call a new sample 'resistant'?

Mortality of pear psylla to amitraz was studied by means of bioassays. Variation between samples, temporal variation within the season in one orchard and spatial variation between Swiss regions were considered. Variation between samples was large enough to produce different Probit functions and LC50 values. Temporal and spatial variations were too small to indicate resistance. Prediction intervals of the pooled functions using bootstrapping were calculated to determine if future samples come from a population with decreased sensitivity. Probabilistic criteria on the population level were proposed for resistance.