| Popis: |
Aim: This study was conducted to model the relationship between discrete dependent variable (yellow stem borer population) and continuous weather variables. Data Description: The yellow stem borer (YSB) population and standard meteorological week (SMW) wise weather variables (temperature, relative humidity, rainfall and sunshine hours) data of Warangal centre (Telangana state) generated under All India Co-Ordinated Rice Improvement Project (AICRIP) from 2013-2021 were considered for the study. The YSB population were recorded daily using light trap with an incandescent bulb and are counted as weekly cumulative catches. Methodology: The weekly cumulative trapped YSB populations and weekly averages of climatological data were considered as inputs to the models under consideration. In this study the classical linear regression i.e. step-wise multiple linear regression and count regression models such as Poisson, negative binomial, zero inflated Poisson and zero inflated negative binomial regression models were employed. Result: The empirical results revealed that the zero inflated count regression models viz., zero inflated Poisson regression and zero inflated negative binomial regression models performed better compared to the classical linear regression, Poisson and negative binomial regression models, further the negative binomial regression model outperformed all models as it yielded lowest mean square error (MSE) and highest R2 values. The average percentage reduction in accuracy of zero-inflated negative binomial regression model over classical model was around 4 percent. Conclusion: Based on the results obtained in this study, it is concluded that the zero inflated models performs better compared to classical models as they are unable to handle the presence of excess zeroes, as a result provides more prediction error and lower R2 values. Further, the models developed in this study will be of great assistance in identifying the factors influencing occurrence of YSB population in rice. |
