Abstract: Outbreaks of the pine caterpillar (Dendrolimus punctatus Walker) of Qinzhou County in Guangxi autonomous region are analysed by means of the double selection stepwise regression method. The density of pine caterpillars and the percentage of infected trees in two months are considered as dependent variables and eighteen factors,for example,the original insect density and the percentage of infected trees,various meteorological factors,seasonal month,control or not,interaction between control and R.H.and others are considered as independent variables.Four critical values are employed for selection,it is found that eliminated independent variables increase with the increase of F_x,thus the reserved independent variables are more important.The results show that the original status of pest population is the most important factor,while the man-made interference (i.e.control or not and interaction between control and R.H.)is another factor of importance.Among various meteorological factors,the sunshine-hour is considered as an important factor and the seasonal month factor should not be neglected. We realize that to forcast the average of pest population density in a larger area is inpossible of accuracy,its meaning is insignificant.It's more reasonable to forcast the infected areas and their respective pest density by our suggested method.