A Comprehensive Modeling Framework for Estimation and Prediction of COVID19 in India

There has been substantial interest worldwide in understanding the current status of Corona Virus Disease (COVID-19) epidemic and prediction of the future path through the pandemic. Many groups are attempting to provide the description of spread and modeling the transmission processes for short and long term projections. Since the epidemic is in its early stage, there is paucity of data for predicting the future course of the disease. The back-calculation approach is one of the methods used in such a situation. The back-calculation reconstructs the past pattern of the infection and predicts the future number of cases with the present infection curve. Lack of information about incubation distribution, effect of intervention on incubation period and errors in reporting the cases lead to uncertainties associated with modeling. This paper attempts to formulate the problem of estimating future COVID-19 cases as estimation of parameters in a multinomial likelihood with unknown sample size by EM algorithm. Illustrations are provided using reported cases in India and discussed.