UTILIZING THE BELL LASSO REGRESSION MODEL FOR RNA-SEQ AND HAEMATOLOGY DATA ANALYSIS

Özet

This article examines the application of the Lasso method for variable selection and regression coefficient shrinkage in the context of the Bell regression model for count data. The primary goal is to deal with multicollinearity, a side effect that occurs when explanatory variables have a high degree of correlation. Under such conditions, parameter estimations are prone to inflation, and the resulting models might not correctly capture the underlying reality. Using a variable selection tactic, penalizing techniques like Lasso can be used to find and eliminate strongly linked variables. In order to address the problem of multicollinearity in count datasets, the current work used the Bell Lasso regression model in conjunction with the Alternative Direction Multiplier Method (ADMM) algorithm. The ADMM algorithm is an effective tool in this context because it is extremely well-suited for addressing optimization problems with complex constraints. In this work, the ADMM algorithm is implemented against the background of the Bell Lasso regression model, and the results derived from practical applications are presented. This paper makes a substantial contribution to the fields of statistical modeling and regression analysis by introducing the ADMM algorithm as a solution for the multicollinearity problem in RNA-Seq (Breast Invasive Carcinoma (TCGA, Firehose Legacy)) and Haematology (four eco-geographic zones from Ghana) count datasets. The outcomes show how well the suggested method works in situations where there is a high degree of correlation between the independent to precisely estimate regression coefficients and choose pertinent variables. Eventually, this research provides insightful knowledge and useful methods for enhancing the interpretability and dependability of regression models in situations when count data exhibits multicollinearity.