Dynamic Modeling in Health Research as a framework for developing statistical applications free of misuse of statistics

TL;DR

This paper presents a new dynamic modeling framework for health research that explicitly captures causal relations, aiming to improve statistical application accuracy and prevent misuse, demonstrated through BMI trend analysis in Finland.

Contribution

It introduces a dynamic modeling framework and the Dynamic Regression Method for health research, emphasizing causal relations and temporal changes in health indicators.

Findings

Dynamic models clarify causal relationships in health data.

The Dynamic Regression Method effectively identifies cohort trends.

Application to Finnish BMI data demonstrates practical utility.

Abstract

We introduce a novel framework for developing statistical applications in health research, based on dynamic modeling of the investigated processes. We formulate the principles of dynamic modeling in health research, which are coherent to those in other fields of research. Dynamic models explicitly describe causal relations which are to be adequately accounted in statistical methods, making them free of misuse of statistics and statistical fallacy. We propose the Dynamic Model of Population Health describing temporal changes in health indicators, having nature of state variables. The Dynamic Regression Method was developed as statistical method for the identification of the model. This method evaluates cohort trends for state variables at each age and calendar year. The method is illustrated by evaluating cohort trends for the Body Mass Index for men, using survey data collected in the years 1982, 1987, 1992, in North Karelia, Finland.