On Huber's Robust Technique and Quantile Regression Models for the Total Production of Field Crops in Oman

In this paper, the problem of modelling the data of the agriculture total production of the Field Crops, agricultural land area, temperature and humidity is studied, the data is collected (for the period 1999-2020) and two different models are developed, i.e. the univariate and multivariate models based on the Huber loss robust technique and quantile regression.A transformed data set is used to mitigate the impact of skewness and stabilize variance.This study evaluates the significance of goodness of fit of the above random variables.Moreover, several assumptions and testing of hypotheses are conducted to identify the behaviour of the data.Due to the size of the collected data, the bootstrap approach is utilized to verify the predictability, goodness of fit, and uniqueness of the estimations.Furthermore, the prediction accuracy of the univariate/multivariate models of Huber loss was resilient when compared to each quantile model, and it was discovered that both models fit the data well.A 95% prediction interval of production is generated and is shown to be valid for the models.Based on real data example, it turned out that the predictive accuracy of robust regression and representative quantile regression models was not significant for univariate/multivariate models.When comparing goodness of fit indices in univariate and multivariate models, the RR and sixth QR models were found to be the best fits for the data.The recommendation, limitations and future research are discussed.