Analysing the dataset

This analysis applies multiple linear regression model selection techniques to the 2020 Behavioral Risk Factor Surveillance System (BRFSS) dataset (n = 5,000), predicting the number of physically unhealthy days in the past 30. Using nine candidate predictors, including mental health days, sleep hours, age, BMI, exercise, general health status, and income, the analysis evaluates model fit using R², Adjusted R², AIC, and BIC. It then walks through best subsets regression, automated selection methods (backward elimination, forward selection, and stepwise), and concludes with an associative model built around sleep hours as the exposure, applying the 10% change-in-estimate rule to systematically identify confounders and arrive at a valid adjusted estimate of the sleep–physical health association.

This study establishes a robust and practical framework for generating a microbially diverse yet stable microcosm system for studying host-microbiota interactions in nematodes. By showing that CME maintains compositional stability during storage and supports reproducible consistent microbial exposure while preserving ecological complexity.