dhs_design(),
mics_design() and steps_design() build the
design from each programme’s standard variable names (DHS: PSU
v021, strata v023, weight v005 /
1e6).survey_prevalence(): design-based
prevalence of any number of binary indicators, overall or by one or more
grouping variables, with design effects.mm_prevalence() now also accepts a
survey design object, supports several by variables and
excludes missing scores by domain estimation (same estimates as before
for complete designs).multimorbidity_index(threshold = )
sets the number of conditions defining multimorbidity (e.g. 3 for
complex multimorbidity) and adds mm_multimorbid.mm_patterns(): prevalence of condition
pairs/triads with observed-to-expected ratios.mm_standardize() and
who_standard_population(): direct age
standardisation to the WHO World Standard Population, with design-based
SEs.mm_bounds(): sensitivity bounds for
“don’t know” answers.survey_concentration_index() gains
type = "generalized", "erreygers" and
"wagstaff", bounds, and by for
group-wise indices.survey_concentration_curve() and
survey_slope_index() (slope and relative
index of inequality).ncd_table(): publication-ready
prevalence tables.autoplot() methods for prevalence,
multimorbidity, concentration-index, concentration-curve and pattern
results.vignette("real-data-nhanes") using real
NHANES 2009-2010 data.survey_map_indicator(),
survey_xgboost(), survey_shap() and
plot_shap_summary() are now labelled experimental.
Concentration-index p-values are computed as
2 * pnorm(-|z|) for accuracy with very small
values.
New simulated example datasets ncd_survey (2,400
adults, STEPS-style raw response codes, stratified two-stage cluster
design with unequal weights) and child_anthro (1,034
children with DHS-coded z-scores), drawn from a known synthetic
population. Generating code is in data-raw/ of the source
repository.
New vignette vignette("full-workflow") demonstrating
every exported function on the example datasets.
survey_concentration_index(): respondents tied on
the ranking variable (for example DHS wealth quintiles) now share the
mid-point of their group’s cumulative-weight interval. Previously ties
were ranked by row order, so the estimate changed when the rows of the
input were reordered. Results for untied ranking variables are
unchanged.
Tests: the replicate-weight validation of the concentration index
now uses a tolerance of 1e-8 (the two calculations agree to machine
precision); added a tie-invariance regression test and tests for
survey_map_indicator() (run when sf is
installed).
Documentation: README status column now distinguishes tested functions from the exploratory machine-learning module, and a Limitations section was added.
Added CONTRIBUTING.md and CITATION.cff;
added the author ORCID.
recode_binary(): converts raw survey coding
(e.g. 1/2/9) to clean 0/1/NA. When no is not specified,
non-yes values are treated as 0.multimorbidity_index(): self-reported multimorbidity
score based on the Functional Comorbidity Index (Groll et al., 2005,
doi:10.1016/j.jclinepi.2004.10.018). Safely handles all-NA
respondents.fci_items(): reference list of 18 standard FCI
conditionsmm_prevalence(): design-weighted population prevalence
with logit confidence intervals via the survey package. Includes lonely
PSU protection.survey_concentration_index(): survey-weighted
concentration index for health inequality analysis with
design-consistent standard errors via Kakwani’s convenient regression
(Kakwani, Wagstaff, & van Doorslaer, 1997,
doi:10.1016/S0304-4076(96)01807-6). Returns full-precision values.who_anthro_score(): WHO severity categorisation for
DHS-style anthropometric z-scores with biologically implausible value
flagging per WHO child growth standards (WHO, 2006).survey_map_indicator(): choropleth mapping of survey
indicators via sfsurvey_xgboost(): case-weighted gradient boosting for
survey data (exploratory)survey_shap(): SHAP feature contributions from
survey_xgboost() modelsplot_shap_summary(): SHAP summary visualisation