inedemogR 0.2.0

This release corrects numerical errors in 0.1.0 identified in a methodological audit and validation of the package against the Human Mortality Database protocol and the official tables of the Spanish statistics institute (INE). Several corrections change numerical results; their effect on every number reported in the accompanying article is tabulated by reproducibility/02_manuscript_numbers.R in the GitHub repository.

Breaking changes and corrected results

• andreev_kingkade_a0() now implements the sex-specific, three-segment Andreev-Kingkade (2015) formulas of HMD Methods Protocol V6, Table 3. 0.1.0 used Coale-Demeny-style coefficients (e.g. female a0 at m0 = 0.004 was 0.064 instead of 0.141).
• Life tables use ax = 0.5 at every age except 0 (0.1.0 used a1 = 0.4); combined-sex a0 is the death-weighted average of the sex-specific values (HMD V6 Eq. 77).
• build_life_tables() replaces the unweighted logit regression with the HMD V6 Kannisto model fitted by Poisson maximum likelihood on deaths and exposures at ages 80-99 (parameters constrained to be non-negative), with the data-dependent replacement age Y (100-deaths rule, 80 <= Y <= 95) and smoothed female-exposure weights for the combined-sex table. It now needs the d_* and e_* columns that compute_death_rates() returns. New arguments kannisto_age and fit_min support sensitivity analysis.
• compute_exposure(): the Lexis correction is (D_L - D_U) / 6 (HMD V6 Eq. 57) with conventional triangle labels; 0.1.0 used (D_U - D_L) / 2 with reversed labels. Only affects deaths supplied with a cohort column; INE deaths use the even split, where the correction is zero.
• European Standard Population 2013 weights sum to 100,000 (95+ = 200); 0.1.0 gave ages 95+ a weight of 600, inflating ASDRs by about 20 per cent.
• Abridged life tables preserve nmx = ndx / nLx in every closed group (Greville nax with the Chiang conversion, constant-hazard fallback); 0.1.0 combined an exponential nqx with a linear nLx.
• decompose_life_expectancy(method = "pollard") integrates the open age interval exactly and closed intervals by the midpoint rule; 0.1.0 treated the open interval as a point and recovered only 22.5 per cent of a change confined to it. Both methods return the attributes e0_difference and residual.
• Fertility follows INE’s conventions: exposure is the mean of consecutive January-1 female stocks; births outside 15-49 are folded into ages 15 and 49 instead of dropped; the mean age at childbearing uses mid-points x + 0.5. asfr and asfr_female are now births per woman (TFR, GRR and NRR are plain sums); asfr_per_1000 is provided for display.

Missing data and validation

• An explicit NA death count stays NA (0.1.0 turned it into a zero rate); a province-year with no death records is excluded rather than treated as zero mortality. Ages absent from INE’s deaths table within a reported province-year remain structural zeros.
• build_life_table() rejects non-contiguous ages and non-finite rates. build_life_tables() withholds province-years with missing exposure (e.g. INE populations top-coded at 85+ before 2002) or too few ages to fit the Kannisto model, and reports them in $failed; $qc is now a tibble.
• validate_life_table() and validate_abridged_life_table() also fail on missing ages, non-finite values, a broken mx = dx / Lx identity, and (full tables) implausible e0; validate_death_rates() checks age coverage. ASDR is NA unless all ages have finite rates.

Reproducibility

• With use_cache = TRUE and cache_dir = NULL, retrieval now caches in a per-session directory under tempdir() (0.1.0 silently did not cache).
• Retrieved objects carry a provenance attribute (request, retrieval time, API, package version).
• New opt-in live integration test (INEDEMOGR_LIVE_TESTS=true) and weekly workflow; reproducibility/ holds frozen inputs, the validation study against INE’s provincial life tables and fertility indicators, and the manuscript-number scripts.

inedemogR 0.1.0

System A: quick multi-geography totals

• get_ine_demog() retrieves live population, births, and deaths totals from INE via ineapir::get_data_table(), at whichever geographic level (municipality or province) each indicator is actually published at.
• get_ine_geo() retrieves municipality/province geometries via mapSpain, with an option to shift the Canary Islands next to the mainland for compact national maps.
• list_ine_indicators() and update_ine_data() round out data discovery and local cache refresh.
• plot_ine_map() provides a highlight-region diagnostic map; general choropleths are covered by System B’s map_indicator() below.

System B: age/sex-disaggregated province data and the mortality pipeline

• get_ine_births(), get_ine_births_by_age(), get_ine_deaths(), and get_ine_population() retrieve province-level, age/sex-disaggregated data directly from INE’s Tempus3 API, following the Spanish subnational Human Mortality Database (SHMD) protocol. get_ine_births_by_age() retrieves single-year age-of-mother birth counts, the input the fertility schedule functions below need.
• compute_exposure(), compute_death_rates(), and build_life_table()/ build_life_tables() implement exposure-to-risk, central death rate (1x1, 5x1, age-standardised), and single-year period life table construction per HMD Methods Protocol V6 (Andreev-Kingkade a0 at age 0, Kannisto old-age smoothing).
• build_abridged_life_table()/build_abridged_life_tables() build standard abridged (5-year age group) period life tables directly from the 5x1 central death rates, complementing the single-year tables above - useful for comparing against other agencies’ published abridged tables.
• download_ine_data() runs any subset of the pipeline and writes results to CSV and/or HMD-format .txt files for use outside R.
• A validate_*() family (validate_population(), validate_births(), validate_births_by_age(), validate_deaths(), validate_deaths_age(), validate_exposure(), validate_death_rates(), validate_life_table(), validate_abridged_life_table()) flags data-quality issues (suppressed cells, non-monotonic life tables, implausible exposure) without failing hard.

Summary demographic indicators

• age_dependency_ratio(), aging_index(), and sex_ratio() compute population-structure indicators from age-disaggregated province data.
• crude_birth_rate(), crude_death_rate(), general_fertility_rate(), infant_mortality_rate(), and rate_of_natural_increase() compute CBR, CDR, GFR, IMR, and RNI.
• age_specific_fertility_rate(), total_fertility_rate(), mean_age_at_childbearing(), gross_reproduction_rate(), and net_reproduction_rate() build the full age-specific fertility schedule (ASFR/TFR/MAC/GRR/NRR) from age-of-mother birth counts - a true total fertility rate, which crude_birth_rate()/ general_fertility_rate() cannot compute on their own.
• birth_death_ratio() computes births per death directly from System A’s totals, needing no age breakdown.
• life_expectancy_summary() and life_expectancy() extract e0/e65 from a period life table.
• decompose_life_expectancy() attributes a difference in life expectancy at birth between two life tables (two provinces, or one province across two years) to age-specific contributions, via Arriaga’s (1984, exact) and Pollard’s (1988, approximate) methods.

Visualization

• plot_population_pyramid() and plot_demog_trend() provide population pyramid and generic indicator time-series charts.
• map_indicator() provides a general-purpose choropleth (binned or continuous) for any geography-keyed tibble; map_life_expectancy() is a thin wrapper bridging the mortality pipeline’s life tables onto province geometry.
• plot_lexis_diagram() draws an age x year mortality surface with birth-cohort diagonals.

Documentation

• vignette("inedemogR-tutorial") is a single comprehensive tutorial covering data retrieval/cleaning/storage (Part I) and demographic analysis/visualization (Part II), with the exact formulas implemented in code and worked examples reproducing the scripts in examples/.