PNADCperiods 0.1.3 (in development on dev branch)

SIDRA access moved to IBGE’s aggregated-data API v3

Between 14 and 16 September 2026, IBGE placed apisidra.ibge.gov.br behind a Cloudflare browser challenge. The host now answers HTTP 403 to most programmatic requests, regardless of the client: ten identical requests measured on 2026-09-18 returned three successes and seven challenges. Every series the package fetches was affected.

fetch_sidra_rolling_quarters() and fetch_monthly_population() now query IBGE’s aggregated-data API v3 (servicodados.ibge.gov.br/api/v3/agregados), which serves the same tables and is not challenged. The legacy SIDRA paths in get_sidra_series_metadata() are unchanged and are translated at request time, so nothing in the public interface moves.

• sidrar is no longer a dependency. Its 0.5.1 fallback to the v3 service rejects whole-series period selections (p/all) and decimal modifiers (/d/), which every path in the metadata table uses. Removing it also drops six transitive dependencies; curl and jsonlite take its place.
• All 90 series were validated against the last bundle fetched through apisidra (2026-09-14): every value over the 569 shared periods is identical, including the fourteen series whose rounding came from a /d/ modifier, which the new client reproduces client-side.
• Two new options: PNADCperiods.sidra_base_url overrides the service host and PNADCperiods.sidra_timeout the per-request timeout (default 60s).

Behaviour on an unreachable API is unchanged: an informative message() and NULL returned invisibly, never a warning() or stop().

Breaking change - pure-cumsum mensalization

mensalize_sidra_series() no longer re-anchors each (Jan, Feb, Mar)-style trio to its rolling-quarter mean as a final step. The internal call to .apply_final_adjustment() was removed from .mensalize_single_series() and from both halves of .mensalize_split_series(). The algorithm now stops at m = y0 + cum: three strictly independent month-position sub-series accumulating from their own starting points.

Why: the legacy final adjustment used rq_lead1/rq_lead2 to enforce mean(m[t-2..t]) == rq[t]. Whenever IBGE published the rolling quarter that ends in month t, the trio (m[t-2], m[t-1], m[t]) would all shift by the same delta rq[t] - mean(y[t-2..t]). In real PNADC series this delta is nonzero (because microdata-derived y0 is not exactly consistent with later-published rq values), so previously published months kept silently moving each time IBGE released a new rolling quarter.

Effect on output: monthly values previously published are stable across IBGE releases. The trio average no longer exactly equals the official rolling-quarter value; the maximum drift observed in 2026-04 real data was about 0.0005% relative to the published rq.

Backward compatibility: the internal helper .apply_final_adjustment() is preserved (still unit-tested and reachable via :::) for users who need to reproduce historical analyses that assumed the legacy behaviour. Calling the public mensalize_sidra_series() no longer invokes it.

PNADCperiods 0.1.2

CRAN policy compliance (response to Brian Ripley, 2026-04-26)

• Internet-using functions now fail gracefully with an informative message() (no warning(), no stop()) and return NULL invisibly when the SIDRA API is unreachable or returns an unexpected schema. Affects fetch_monthly_population(), fetch_sidra_rolling_quarters(), and pnadc_apply_periods() (which falls back to returning the data with the crosswalk applied but no calibrated weights when SIDRA is unavailable).
• New mocked-API tests (testthat::local_mocked_bindings) verify the graceful-failure behaviour offline.
• Test suite no longer depends on the SIDRA API: 22 calibration and pipeline tests now use locally-constructed target_totals mocks instead of fetching live SIDRA population data.

Bug fixes — mensalize_sidra_series()

• Trailing NA in the rolling-quarter input (e.g., when IBGE has published IPCA/INPC for a month but not yet the corresponding PNADC rolling quarter) no longer produces phantom mensalized values that silently echoed the value of the same mesnotrim position three months earlier. The mask now zeros out only trailing positions after the last observed rq, leaving leading positions intact.
• IPCA lag is pre-computed before the PNADC filter is applied, removing contamination of starting points by future-IPCA observations.
• INPC lag is pre-computed in compute_series_starting_points(), with test coverage for derived-series propagation through aggregates.

Documentation

• README, vignettes and pkgdown navbar updated to recommend install.packages("PNADCperiods") now that the package is on CRAN.

PNADCperiods 0.1.1

CRAN resubmission

• Expanded all acronyms in DESCRIPTION (IBGE, SIDRA, UPA, API)
• Added Hecksher (2020) reference with URL
• Replaced \dontrun{} with \donttest{} for API examples; created runnable examples for validate_pnadc() and get_sidra_series_metadata()
• Fixed missing anchor parameter in compute_starting_points_from_microdata()
• Removed .onLoad that globally set data.table thread count
• Added skip_on_cran() to all API-dependent tests
• Added fst to Suggests

PNADCperiods 0.1.0

Initial CRAN release

Microdata period identification

• pnadc_identify_periods(): Build crosswalk mapping survey observations to months, fortnights, and weeks using IBGE’s technical break rules and respondent birthdates.
• pnadc_apply_periods(): Apply crosswalk and calibrate weights to SIDRA population totals with adaptive hierarchical cell definitions.
• pnadc_experimental_periods(): Probabilistic assignment and UPA aggregation strategies for improved fortnight/week determination rates.
• validate_pnadc(): Input validation for required PNADC variables.

SIDRA mensalization module

• get_sidra_series_metadata(): Metadata for 86+ PNADC rolling quarter series available from the IBGE SIDRA API.
• fetch_sidra_rolling_quarters(): Download rolling quarter data from SIDRA.
• mensalize_sidra_series(): Convert rolling quarter series into exact monthly estimates via matrix inversion, without requiring microdata access.
• fetch_monthly_population(): Fetch monthly population totals from SIDRA.
• compute_starting_points_from_microdata(): Compute y0 starting points for the mensalization algorithm from survey microdata.