Exploring The Factors Related To The Yield Of Sunflower Crop: An Application Of Robust And Ridge Regression Analysis

Abstract

The first production of sunflower was introduced to Pakistan in 1960’s and was one of the country primary oilseed crops. According to Food Agriculture Organization (FAO), 152,675 hectares of sunflower were cultivated in Pakistan in 2014 and a total of 3240 tons seeds were produced. Pakistan is yielding 0.14 million tons production of sunflower only in the world. There are many factors that contribute to increase yield per hectare, but fertilizer management is more important to enhance sunflower crop growth development and achene (one-seeded dry fruit) production. The current study focus on selection of efficient model for sunflower production in Pakistan. For this purpose, sunflower dataset is taken for analysis. The yield is taken as dependent variable and all other factors such as village name, plant population, soil type ,quantity of seed, urea, Dap, No water well / Tube well, usage of machine, weight residual, last crop, seed treated, Attack pest and harvest price are taken as independent variables. Three steps are used to select an efficient model. In the first step, Correlation matrix and box Plots are used to analyze the multicollinearity ¹and outliers respectively. No multicollinearity among predictors is deducted while the boxplot reveals that there are outliers present in the dataset. In the second steps, due to the presence of outliers in the datasets, Robust Regression will be used for the purpose of analysis. Three M-estimators are used of robust regression (Huber M- estimator, Hampel M-estimator and Tukey bisquare M-estimator). Final steps were considered on the basis of efficient model selection by using model selection criteria such as Mean square error (MSE), Mean Absolute Percentage Error (MAPE), Akaike’s information criteria (AIC) and Bayesian information criteria (BIC). The efficient model for “yield of sunflower” is selected by Hampel M estimator and is preferred on the basis of minimum value of MSE, MAPE, AIC and BIC.