TIME SERIES ANALYSIS OF RAINFALL USING HETEROSKEDASTICITY MODELS
By
Rashmi Bhardwaj and Varsha Duhoon
University School of Basic and Applied Sciences, Non-Linear Dynamics Research Lab, Guru
Gobind Singh Indraprastha University, Dwarka, Delhi-110078, India.
Email:rashmib22@gmail.com, varshaduhoon5@gmail.com
(Received : February 04, 2020 ; Revised: June 21, 2020)
DOI: https://doi.org/10.58250/jnanabha.2020.50123
Abstract
Weather forecasting is predicting present state of weather by the help of analyzing collected data such as temperature, humidity, wind etc. to analyze atmospheric processes and determine how weather condition is going to change in future. Weather forecasting is important not only for prediction but also to prepare for future coming events if any such as cyclone, heavy rainfall, hails which can cause harm to the agricultural production of the country and hence affect the livelihood of the farmers. Continuous change in variance of time series over time; such a process is termed as Volatility. Heteroskecdascity refers to increasing variance in a way, such as increase in trend, this property of series is termed as Heteroskecdascity. Objective of the paper is to analyze, model and predict Rainfall time series of Delhi region from January 01, 2017 to December 31, 2018 using Heteroskecdascity model such as ARCH, GARCH, TARCH/ GJR-GARCH, EGARCH models to select most suited model on basis of probability value of the model hence calculated. Further, to analyze and choose model has full filled required conditions the model is checked for Serial Correlation, ARCH Effect, Normal distribution of Residuals, ARCH-LM test is applied, AIC, SIC values are calculated. GJR-GARCH model is most suited model among all models tested for modeling and analyzing rainfall. Model selection is done based on AIC value and SIC value calculated.
2010 Mathematics Subject Classifications: 33C60, 33C45.
Keywords and phrases: Volatility, Heteroscedascity, ARCH, GARCH, TARCH, EGARCH, Time Series, Rainfall.