Statistical Analysis Of The Incubation Period Of COVID-19 Based On The Generalized Odds Rate Mixture Cure Model

Since the epidemic of COVID-19,many scholars have discussed and studied its incubation period,and speculated on the length of the virus incubation period.In this article,asymptomatic patients are included in the research objects.Regard the incubation period of COVID-19 as interval-censored data with a cure-function and apply the generalized odds rate mixture cure model to do statistical analysis.The model is flexible enough due to the existence of a transformation parameter,including kinds of survival models in the latency part such as the proportional hazards model and the proportional odds model.The transformation parameter is selected according to the log-likelihood values,which can provide a reference for model selection in practice.In order to solve the likelihood function better,a gamma-Poisson latent variable is introduced to expand the likelihood function.This method greatly simplifies the structure of the likelihood function.Numerical simulations under different interval censored rates show that the method based on the expanded likelihood function in terms of parameter estimates,average estimated standard errors and convergence time is significantly better than directly maximizing the observed likelihood function.The model is applied to the analysis of the incubation period of COVID-19 to estimate parameters,determine risk factors and predict new individuals,and is compared with other cure models.Based on the data used in this article,the model corresponding to the maximum log-likelihood is exactly the proportional hazards mixture cure model.Comparing the proportional odds mixture cure model and the generalized proportional odds mixture cure model,it can be found that the values of the transform parameter affect the judgment of risk factors,so it is important to choose an appropriate cure model to fit the survival data.