Age-gender-country-specific death rates modelling and forecasting: a linear mixed-effects model

Varování

Publikace nespadá pod Ústav výpočetní techniky, ale pod Přírodovědeckou fakultu. Oficiální stránka publikace je na webu muni.cz.
Autoři

DASTRANJ Reza KOLÁŘ Martin

Rok publikování 2025
Druh Článek v odborném periodiku
Časopis / Zdroj Scandinavian Actuarial Journal
Fakulta / Pracoviště MU

Přírodovědecká fakulta

Citace
www https://doi.org/10.1080/03461238.2025.2492181
Doi http://dx.doi.org/10.1080/03461238.2025.2492181
Klíčová slova Life insurance; mortality forecasting; restricted maximum likelihood; model selection; random walks with drift
Přiložené soubory
Popis A linear mixed-effects (LME) model is proposed for modelling and forecasting multi-population age-specific death rates (ASDRs). The innovative approach that we take in this study treats age, the interaction between gender and age, their interactions with predictors, and cohort as fixed effects. Furthermore, we incorporate additional random effects to account for variations in the intercept, predictor coefficients, and cohort effects among different age groups of females and males across various countries. We will show that the LME model is identifiable. We will use data from the Human Mortality Database (HMD) to illustrate the procedure. We will assess the predictive performance of the LME model in comparison to the Lee-Carter (LC) models fitted to individual populations. Additionally, we evaluate the predictive accuracy of the LME model relative to the Li-Lee (LL) model. Our results indicate that the LME model provides a more precise representation of observed mortality rates within the HMD, demonstrates robustness in calibration rate selection, and exhibits superior performance when contrasted with the LC and LL models.
Související projekty:

Používáte starou verzi internetového prohlížeče. Doporučujeme aktualizovat Váš prohlížeč na nejnovější verzi.

Další info