The Role of Microsimulation in Longitudinal Data Analysis

Abstract

Abstract:Microsimulation is well known as a tool for static analysis of tax and transfer policies, for the generation of programmatic cost estimates, and dynamic analyses of socio-economic and demographic systems.However, microsimulation also has the potential to contribute to longitudinal data analysis in several ways, including extending the range of outputs generated by a model, addressing several defective-data problems, and serving as a vehicle for missing-data imputation. This paper discusses microsimulation procedures suitable for several commonly-used statistical models applied to longitudinal data. It also addresses the unique role that can be played by microsimulation in longitudinal data analysis, and the problem of accounting for the several sources of variability associated with microsimulation procedures. Methodological Issues -Douglas A. Wolf 314 RésuméLa micro-simulation est bien connu comme un outil pour l'analyse statique des politiques d'impôts et de transfert, pour l'estimation des coûts des programmes, et pour l'analyse dynamique des systèmes socio-économiques et démographiques. Cependant, le micro-simulation offre également des possibilités intéressantes pour l'analyse de données longitudinale, y compris le développement d'autres produites du modèle, la possibilité d'adresser des problèmes de données déficientes, et comme véhicule pour l'imputation de données manquantes. Cet article discute des procédures de micro-simulation appropriées à plusieurs modèles statistiques qui utilisent des données longitudinales. Il adresse également le rôle unique qui peut être joué par la micro-simulation dans l'analyse de données longitudinale, et le problème posé par les multiples sources de variabilité dans les procédures de micro-simulation.

Uncertainty Analysis in Population-Based Disease Microsimulation Models

“…However, systematic presentations of UA for predictive PMS models with multiple outcomes are lacking. In particular, there has been little discussion of UA in the context of chronic disease models with epidemiological or burden-of-disease outcomes, such as those predicting the incidence and prevalence of a disease in a population, its impact on mortality and quality of life, or its costs [3,12]. In the following section we provide an overview of the key steps in UA for such models.…”

“…Other types of uncertainty, including uncertainty in the starting population [12] and structural model uncertainty [16], have been studied. Structural uncertainty is associated with the choice of model structure, including the statistical models used, variables in the model, and the specified relationships between the variables [31].…”

“…A common sampling technique is to generate a uniform random number (between 0 and 1) through one of the existing random number generator algorithms [61] (e.g., pseudorandom number generators are among the most used methods that deterministically perform series of tasks to generate statistically sound (uniform) random numbers [12]). Then, using the inverse of the cumulative density function (CDF), known as the quantile function, one finds a real value of the parameter.…”

“…The uncertainty associated with any given outcome in a disease simulation model is typically a reflection of the complex interplay among the model variables [12]. Due to this complexity, estimating uncertainty with analytical approaches is often impossible, and numerical methods, such as the Monte Carlo (MC) method, are almost always required.…”

“…Population-based models differ from models commonly used in cohort-based cost-effectiveness studies that model a single cohort of patients [9]. Unlike macrolevel simulation models (e.g., cell-based [10] or compartmental models [11]), microsimulation (MS) models generate a life history for every individual in a population [12,13] and provide population-level outcomes by aggregating the individuals' event histories [14,15].…”