Changelog • spi

spi (development version)

spi_case_date(), spi_case_panel(), spi_pop_panel(), spi_pop_index(), spi_window() and spi_panel_months() turn POLIS AFP and population tables into the inputs of spi_expected().
spi_window_summary() reads SPI over a rolling window from one fit on the full history, for districts or for provinces summed within each draw. spi_index() is unchanged and still refits per assessment year.
spi_window_classify() sets a rolling-window SPI against its threshold, with the 90% interval rule, and the NPAFP rate against its target. Units with an expected count below min_expected are left unclassified.
spi_rolling_index() chains these steps, fitting the model when no fit is given.

spi_index() centres each posterior draw on the national ratio from the same draw, then summarises. Uncertainty shared by all districts in a period no longer widens district intervals. Recalculate centred results from 0.2.0.
Simple SPI reports explain missing SPI values when the national ratio is zero.

spi_simple() calculates SPI from each district’s earlier NPAFP rate, without INLA or boundaries.
spi_simple_explain() shows one district’s simple calculation, in English or another language.
spi_compare_npafp() and spi_field_guide() accept simple SPI results.
spi_check_inputs(method = "simple") checks simple SPI inputs.

New articles on the simple SPI, the model-based SPI, data preparation and POLIS data. Simplified language throughout.

Renamed the package from blindspot to spi and the bs_ prefix to spi_. Recreate saved results.
spi_concordance() is now spi_compare_npafp(), with renamed categories and columns.
bs_spi_prospective() is now spi_index(). bs_spi() and the spi_triangulate*() functions are removed.

National centring is the default, with ratios in $national.
spi_check_inputs() checks inputs before fitting.
spi_field_guide_pager() creates printable district reports.
STEPS covers strength, timeliness, extent, persistence and stool adequacy.
year_end_month supports reporting years ending in any month.
Model defaults are now overdispersion = "nb" and year_effect = "iid". seed controls both R and INLA sampling.

Fixed NPAFP populations for custom reporting years, input validation and missing-name handling in reports.