重回帰分析によるヤノネカイガラムシ雌成虫寄生数の予察(第2報)

概要

論文の詳細を見る
Multiple regression models for predicting the arrowhead scale population densities were designed 2 years ago. These models were further improved by incorporating the data accumulated at the experiment stations of Kanagawa, Shizuoka, Hiroshima and Kumamoto during 5 years after the previous model had been constructed. The following 3 years of models of comprised statistically and/or biologically significant variables. Type A : involved the stepwise forward regression method based on the smallest prediction sum of squares ; Type B : the same method was used based on the smallest prediction sum of squares ; Type C : included fewer variables selected on the basis of biological of biogical consideration or by intuition. Regression coefficients were determined by using the following 2 kinds of data sets : (i) each of 4 sets collected at the 4 respective experiment stations and (ii) the pooled data of the 4 experiment stations. Compared with the previous models, the multiple correlation coefficients in the new models were a little smaller, although they reached a value of 0.75 and the multiple regression coefficients were more stable among the 4 stations than in the previous models. These new models were used for predicting the arrowhead scale population density in the year following the observations for model building . The results show that Type C model offered the highest predictability and Type B was as good as Type C. Predictability was high even in the case of the model for the respective station using their own data, but type A was not particularly adequate. Conclusions are as follows : if the sample size for constructing a regression model is adequate, reliable multiple regression coefficients can be obtained, and such multiple regression type models, especially based on the prediction sum of squares, are reliable enough for forecasting of population density.
日本応用動物昆虫学会の論文
1981-02-25

重回帰分析によるヤノネカイガラムシ雌成虫寄生数の予察(第2報)

スポンサーリンク

概要

著者

関連論文

スポンサーリンク