Skip to contents

Every plot takes an sdid_run from fit_spec() (or load_run()) and returns a ggplot object, so it can be themed, saved or combined further.

Usage

plot_sdid_trajectory(run, se_method = "none")

plot_gap_path(run)

plot_event_study(run)

plot_donor_weights(run, top_n = 20L, material = 0.01)

plot_lambda_weights(run)

plot_sdid_units(run)

plot_placebo_distribution(run, bins = 50)

plot_placebo_trim(run, bins = 50)

plot_donor_influence(run, label_n = 6L)

plot_scm_gaps(run, corrected = FALSE)

Arguments

run

An sdid_run.

se_method

Passed to synthdid::synthdid_plot().

top_n

Donors to show.

material

Weight marked as the materiality threshold.

bins

Histogram bins.

label_n

Donors to label.

corrected

For plot_scm_gaps(): plot bias-corrected gaps instead of the uncorrected ones.

Value

A ggplot.

Details

  • plot_sdid_trajectory() – treated unit and synthetic control over time, with the time weights along the bottom (synthdid's own plot, restyled).

  • plot_gap_path() – treated minus synthetic in every period, measured against the baseline the estimator subtracts, so the post-period points average exactly to the ATT.

  • plot_event_study() – the effect in each post-treatment period.

  • plot_donor_weights(), plot_lambda_weights() – unit and time weights.

  • plot_sdid_units() – each donor's contribution (synthdid's units plot).

  • plot_placebo_distribution(), plot_placebo_trim() – the placebo reference distribution behind the standard error, and where trimming cut it.

  • plot_donor_influence() – each donor's share of the estimate against its share of the standard error (see donor_influence()).

  • plot_scm_gaps() – for estimator = "scm": the treated unit's gap against every placebo's, the classic permutation figure.

Plots that need the fitted object or the placebo draws require a run fitted with keep = "all" (the default).

Examples

# \donttest{
run <- fit_spec(sdid_spec(inference = list(replications = 50)), prop99_panel(), seed = 1)
plot_sdid_trajectory(run)

plot_gap_path(run)

plot_placebo_distribution(run)

# }