This paper describes a large scale application of an earlier proposed
strategy, based on statistical design and OSAR modelling, for screenin
g and ranking hazardous chemicals. Goal of this application has been t
o define a model for estimating the rate constant of the tropospheric
reaction of hydrohalofluoro-, -chloro-, and -bromo-methanes, -ethanes,
and -propanes with the OH radical (k(OH)) The addressed compounds wer
e 390, and they were characterised with 11 variables drawn from the st
ructural formula. The model was developed using a training set of 16 h
ydrohalocarbons, selected out by a D-optimal design from the 55 compou
nds for which experimental k(OH)'s were available. The predictive capa
city of the resulting model, measured in terms of the parameter lower
case (Standard Deviation of the Errors of Predictions) was 0.17 (with
log k(OH) ranging, from -12.04 to -14.72). These data are confirmed <<
experimentally>> by comparing the predicted with the actual log k(OH)
of an <<external>> validation set of compounds. Finally, the model was
tentatively expanded to include the above-mentioned hydrohalo alkanes
plus their fully halogenated homologues.