If you are setting or reviewing base mortality for defined benefit pension, annuity or life assurance liabilities then logmu is designed for you1, with state of the art tooling, including support for postcode-based models, and performance at scale.
The logmu feature set:
Flexible mortality framework incorporating time-based covariates and arbitrary proportional hazards models.
A/E analytics including confidence intervals and residuals, with visualisation.
Fit proportional hazards models by maximum likelihood, select between them by AIC, and cluster categorical covariates — all in a single pipeline.
Customisation — weighting (e.g. amounts vs lives), statistical relevance (e.g. treat older data as less reliable) and time-based inclusion criteria (e.g. sub-setting experience by age) are built into every function.
Blazingly fast – logmu takes full advantage of SIMD vectorisation and multi-threading. A/E for XX thousand records takes under Y seconds.
The principles underlying logmu have been applied by the author to longevity transactions totalling c.£100Bn across the UK, US and Netherlands.
A typical A/E analysis — data to charts — takes under 20 lines of R:
# exp_data: experience dataset with fields sex (character) and pension (double)
S4PMA <- mortality_table(...)
S4PFA_110 <- mortality_table(...)
mortality <- mortality(if (.i$sex == "male") mort_m else mort_f)
includes <- aged(60, by = 5, to = 100) | periods(2010, by = 5, to = 2025)
by_lives <- aev(exp_data, mortality, includes, weight = .i$pension > 0)
by_amounts <- aev(exp_data, mortality, includes, weight = .i$pension)
plot(by_lives)
plot(by_amounts)