Statistical issues of regression analysis on development of an age-predictive equation

Area of Science:

• Gerontology
• Biostatistics
• Exercise Physiology

Background:

• Accurate age prediction is crucial for epidemiological and clinical studies.
• Physical fitness and body composition can be indicators of biological aging.
• Regression analysis offers a statistical framework for developing predictive models.

Purpose of the Study:

• To establish a predictive equation for chronological age based on physical fitness measurements.
• To identify the most significant physical fitness variables for age prediction.
• To evaluate the performance of different regression models using statistical criteria.

Main Methods:

• Regression analysis was employed to model the relationship between age and physical fitness variables.
• Statistical information criteria, including Akaike's information criterion, were used for model selection.
• Variables analyzed included physical fitness, body composition, blood pressure, and bone strength.
• Residual examinations and multi-collinearity assessments were conducted.

Main Results:

• The optimal predictive equation for age was determined as: ŷ = 90.0 -1.294 * SITUP - 0.467 * MAP - 0.094 * BALANCE (R = 0.837).
• This model utilized sit-ups, mean arterial pressure (MAP), and balance as predictors.
• Including all ten variables initially yielded a higher R-value (0.864), with the adjusted R-squared peaking at the fifth step.

Conclusions:

• Physical fitness measurements, particularly sit-ups, mean arterial pressure, and balance, can effectively predict age in Japanese males.
• Akaike's information criterion is a valuable tool for selecting parsimonious and accurate predictive models.
• The findings suggest a link between physical condition and biological aging processes.