Main analysis function

Usage

run_scd_analysis(data, options = list())

Arguments

data: Data frame containing SCD data with columns case, phase, and outcome.
options: Named list of analysis options. Recognized entries: formula (model formula or string; defaults to a design-based piecewise formula), scd_type (design passed to process_scd_phases(); defaults to detect_design_type()), family (brms family name, default "gaussian"; "binomial" needs a numeric trials column, see Details), prior_type ("default", "informative", "meta_analytic", or "sensitivity"), prior_effect_mean/prior_effect_sd (for the meta-analytic prior), intervention_goal ("increase" or "decrease", passed to the posterior decision summary), expected_effect_size, quick_test, prior_predictive (logical; when TRUE adds a prior predictive check to the diagnostics), and exposure (name of a positive exposure column for the "poisson", "negbinomial", "zero_inflated_poisson", or "zero_inflated_negbinomial" family; adds offset(log(<exposure>)) through add_exposure_offset()).

Value

An object of class scd_analysis containing data summary, power estimate, fitted model, effect sizes, a posterior decision summary (decisions), and diagnostics.

Details

The decision summary reports a ROPE probability for level-change terms only, with the region of default_rope(). Slope terms (trend_*, time_*) get NA: a per-session slope is negligible or not only relative to the number of sessions it is projected over.

Proportions of a known number of trials or intervals can be modeled with family = "binomial" when the data carry a numeric trials, n_trials, opportunities, or intervals column; the response is then outcome | trials(<col>). The beta family stops on outcomes at 0 or 1 and points to the (y * (n - 1) + 0.5) / n squeeze or the binomial family.

A formula with an ar() term warns when time differences within a series are not constant or when outcomes are missing: brms lags by row order, while an AR(1) model assumes equidistant occasions, and dropping missed sessions alters the series.

Examples

if (interactive()) {
run_scd_analysis(simulate_scd_data())
}