Statistical age-period-cohort analysis: a review and critique

Area of Science:

• Epidemiology
• Biostatistics
• Demography

Background:

• Age-period-cohort (APC) analysis is widely used to study disease rates and population trends.
• Statistical modeling of APC data commonly employs the multiple classification model.
• This model accounts for age, period, and birth cohort effects.

Purpose of the Study:

• To discuss the identifiability problem in APC statistical modeling.
• To illustrate the adverse effects of this problem on APC analysis results.
• To review potential issues with two-factor models and alternative approaches.

Main Methods:

• Descriptive and statistical age-period-cohort (APC) analysis.
• Numerical illustration of the identifiability problem's impact.
• Discussion of various modeling approaches and their limitations.

Main Results:

• The multiple classification model for APC data suffers from an inherent identifiability problem.
• This problem can lead to inaccurate or misleading results in APC modeling.
• Limitations exist in two-factor models and other current APC modeling strategies.

Conclusions:

• The identifiability problem poses significant challenges for accurate APC statistical modeling.
• Interpretational limitations arise from the intrinsic characteristics of APC datasets.
• The current state of statistical modeling for APC data is considered to be in an early developmental stage.