Exploring and validating different relationships among various biomarkers by using both linear and nonlinear, single variable and multiple variables regression analysis models and collected big data of a type 2 diabetes patient based on GH-Method: math-physical medicine (No. 549)

Abstract

Gerald C Hsu

In the author’s previous medical research reports, he mainly applied physics theories, engineering models, mathematical equations, computer big data analytics and artificial intelligence (AI) techniques, as well as some statistical approaches to explore and interpret various biophysical phenomena. However, the majority of medical research papers he has read thus far are primarily based on statistics. As a result, for his paper no. 540 through no. 548 (except no. 545), he has selected some basic statistical tools, such as correlation, variance, p-values, and regression analyses, to study his various biomarkers using linear regression analysis model with either single variable or multiple variables. In this particular paper, he has selected 5 cases to compare their concluding findings using both linear regression and nonlinear regression. The nonlinear regression models include exponential, logarithmic, polynomial, and power. The biomarkers he selected to study are body weight for obesity, FPG & PPG for diabetes, and CVD/Stroke risk probability (chronic disease complication). The inputs or independent variables are carbohydrates & sugar intake amount, body weight, 4 medical conditions score, 6 lifestyle details score, sleep, food consumption quantity, and HbA1C to indicate the insulin resistance level. Depending on his selected case, the body weight has served as either output of dependent variable (symptom) or input independent variable (cause). By the way, since 1/1/2012, the author has collected ~3 million data thus far regarding his health, lifestyle, organs, and diseases.

PDF