Tianyi Guan

About

This project investigated demographic, behavioural, and health-related factors associated with systolic blood pressure using multivariable statistical modelling. A dataset of 500 individuals was analyzed across lifestyle, anthropometric, socioeconomic, and treatment-related variables, with the objective of developing a parsimonious and interpretable regression model for SBP. The analysis combined exploratory data assessment, variable-dependence screening, nested model comparison, interaction modelling, and residual diagnostics to identify factors that contributed meaningfully to variation in blood pressure.

Research Focus

To identify factors independently associated with systolic blood pressure and determine whether interactions among lifestyle, anthropometric, and treatment-related variables improve explanatory performance beyond a main-effects regression model.

Study Design

A cross-sectional dataset containing 500 observations and 18 variables was analyzed, with systolic blood pressure as the continuous response. Exploratory analysis included distributional assessment, correlation analysis among continuous variables, and chi-square tests of independence among categorical predictors. Redundant variables were excluded where substantial structural overlap was present, including weight and height in relation to BMI and gender in relation to childbearing-potential coding. A complete multiple linear regression model was progressively reduced using nested F-tests, after which biologically and statistically relevant interaction terms were evaluated. The final model was assessed using residual normality, homoscedasticity, independence, leverage, and influence diagnostics.

Key Findings

The final parsimonious model retained alcohol use, BMI, hypertension treatment, smoking, exercise level, and several interaction terms, and was highly significant overall (F = 11.37, p < 2.2 · 10^-16). It explained approximately 21.3% of adjusted SBP variability, improving on the approximately 17.8% adjusted variance explained by the initial full model. Higher BMI, smoking, and high alcohol use were associated with higher SBP, whereas high exercise level was associated with lower SBP. Treatment-related interaction terms further indicated that several predictor–SBP relationships varied according to hypertension-treatment status. These results were interpreted as associations rather than causal effects.

Key Methods

Multiple linear regression; exploratory data analysis; correlation analysis; chi-square testing; nested F-tests; interaction modelling; model reduction; adjusted R² comparison; residual diagnostics; leverage and influence assessment;

Keywords

Biostatistics; systolic blood pressure; multiple linear regression; cardiovascular risk factors; BMI; smoking; alcohol use; physical activity; interaction effects; model selection; regression diagnostics;

Project Status

Completed (2025)