Statistical Analysis Of Varying-coefficient Partially Functional Linear Quantile Regression Models Based On Laryngoscope Image And Censored Data

Laryngoscope image dataset refers to the collection of human laryngeal image data captured using a laryngoscope.These images are usually used for medical research and diagnosis,to determine the degree of laryngeal lesions.Laryngoscopy image datasets are usually collected by professional doctors or medical researchers,including normal laryngeal images and images of various laryngeal diseases,such as vocal cord polyps,laryngeal cancer,etc.Exploring the impact of laryngoscopy images and some scalar information on the degree of laryngeal lesions,and considering the differences in influence between different individuals,becomes even more challenging.This paper will use the variable coefficient partial function type linear quantile regression model to fit,to analyze the factors affecting the time of laryngeal carcinogenesis.The specific contents are as follows:The first part estimates the parameters of the variable coefficient partial functional linear quantile regression model for the right censored data.First,the linear quantile regression model with variable coefficient partial function is established.The slope function is approximated by FPCA method,and the variable coefficient function is approximated by regression spline method.The complex form of the linear quantile regression model with variable coefficient partial function is simplified into the form of ordinary linear regression model.Due to the presence of censored data,the inverse censored probability weighting method was used to estimate unknown parameters,and simulation studies were conducted to demonstrate the large sample nature of the estimation.The results were applied to laryngoscopy image data to analyze the locations and other influencing factors that affect the time of laryngeal cancer transformation.The second part selects variables for the multidimensional variable coefficient partial functional linear quantile regression model based on the right censored data.On the basis of the model established in the first part,considering that the functional covariates are multidimensional,the functional covariates in the variable coefficient partial functional linear quantile regression model are selected,and the Group Alasso and Group SCAD methods are compared.Finally,in simulation studies,it was demonstrated that the Group SCAD method had better variable selection performance and was applied to laryngoscopy image data to screen out areas that had an impact on the time of laryngeal cancer transformation.