Package index

Overview

weightflow weightflow-package

weightflow: declarative survey weighting

Build and run a recipe

Define the recipe, estimate it, and pull the weights out.

weighting_spec()

Start a weighting specification

prep()

Estimate the weighting cascade

collect_weights()

Extract the data with the computed weights

y_model()

Specify a working model for a study variable y

Adjustment steps

The staged adjustments, applied in the order you pipe them.

step_unknown_eligibility()

Unknown-eligibility adjustment

step_drop_ineligible()

Drop ineligible (out-of-scope) units

step_select_within()

Within-household selection adjustment

step_nonresponse()

Nonresponse adjustment

step_calibrate()

Calibration to population totals

step_model_calibration()

Model calibration (model-assisted, Wu & Sitter 2001)

step_trim()

Trim extreme weights

step_trim_weights()

Automatic weight trimming (survey-style)

step_round()

Round the final weights

step_rescale()

Rescale (normalize) the weights

step_assert()

Assert conditions on the weights at this point of the cascade

Diagnostics and reporting

Inspect, summarise and report the cascade.

summary(<prepped_weighting_spec>)

Detailed per-step diagnostics

plot(<prepped_weighting_spec>)

Diagnostic plots for the weights

weight_factors()

Per-unit adjustment factors table

design_effect()

Kish design effect from unequal weighting

report_weighting()

Build a nice HTML report of the weighting recipe

Variance estimation

Bootstrap that re-applies the recipe, plus survey/srvyr bridges.

bootstrap_weights()

Bootstrap replicate weights that re-apply the recipe

bootstrap_estimate() boot_total() boot_mean()

Bootstrap estimate, standard error and confidence interval

as_svydesign() as_svrepdesign()

Export weightflow weights to a survey design

collect_replicate_weights()

Collect replicate weights into a data frame ready for srvyr

Example data

population

Example target population for weightflow

sample_survey

Example survey sample (take-all roster)

sample_one

Example survey sample (select-one-person, multistage)