SaturatedEventStudy

SaturatedEventStudy(
    data,
    yname,
    idname,
    tname,
    gname,
    att=True,
    cluster=None,
    xfml=None,
    display_warning=True,
)

Saturated event study with cohort-specific effect curves.

Attributes

Name	Type	Description
data	pd.DataFrame	Dataframe containing the data.
yname	str	Name of the outcome variable.
idname	str	Name of the unit identifier variable.
tname	str	Name of the time variable.
gname	str	Name of the treatment variable.
cluster	str	The name of the cluster variable.
xfml	str	Additional covariates to include in the model.
att	bool	Whether to use the average treatment effect.
display_warning	bool	Whether to display (some) warning messages.

Notes

Fits a fully saturated event study that interacts every treatment cohort with every relative period. Each cohort has its own set of event time effects, which avoids the comparisons between already-treated units that bias static two-way fixed effects under heterogeneous treatment effects (Sun and Abraham 2021, Journal of Econometrics).

Examples

Returned by event_study() with estimator="saturated". aggregate() aggregates the cohort-specific effects by event time.

import pyfixest as pf

data = pf.get_motherhood_event_study_data()

fit = pf.event_study(
    data,
    yname="log_earnings",
    idname="unit",
    tname="year",
    gname="g",
    estimator="saturated",
)
fit.aggregate().head()

	Estimate	Std. Error	t value	Pr(>\|t\|)	2.5%	97.5%
period
-17.0	-0.528314	0.012096	-43.677375	0.0	-0.552022	-0.504607
-16.0	-0.311872	0.067012	-4.653967	0.000003	-0.443213	-0.180531
-15.0	-0.066293	0.022019	-3.010666	0.002607	-0.109449	-0.023136
-14.0	0.004688	0.016255	0.288419	0.773026	-0.027172	0.036548
-13.0	0.035952	0.013157	2.732575	0.006284	0.010165	0.06174

iplot_aggregate() plots the aggregated effects.

fit.iplot_aggregate()

Methods

Name	Description
SaturatedEventStudy.aggregate	Aggregate the fully interacted event study estimates by relative time, cohort, and time.
SaturatedEventStudy.estimate	Estimate the model.
SaturatedEventStudy.iplot	Plot DID estimates.
SaturatedEventStudy.iplot_aggregate	Plot the aggregated estimates.
SaturatedEventStudy.summary	Get summary table.
SaturatedEventStudy.test_treatment_heterogeneity	Test for treatment heterogeneity in the event study design.
SaturatedEventStudy.tidy	Tidy result dataframe.
SaturatedEventStudy.vcov	Get the covariance matrix.

SaturatedEventStudy.aggregate

aggregate(agg='period', weighting='shares')

Aggregate the fully interacted event study estimates by relative time, cohort, and time.

Parameters

Name	Type	Description	Default
agg	str	The type of aggregation to perform. Can be either “att” or “cohort” or “period”. Default is “att”. If “att”, computes the average treatment effect on the treated. If “cohort”, computes the average treatment effect by cohort. If “period”, computes the average treatment effect by period.	`'period'`
weighting	str	The type of weighting to use. Can be either ‘shares’ or ‘variance’.	`'shares'`

Returns

Name	Type	Description
	pd.Series	A Series containing the aggregated estimates.

SaturatedEventStudy.estimate

estimate()

Estimate the model.

Returns

Name	Type	Description
	Feols	The fitted Feols model object.

SaturatedEventStudy.iplot

iplot()

Plot DID estimates.

SaturatedEventStudy.iplot_aggregate

iplot_aggregate(agg='period', weighting='shares')

Plot the aggregated estimates.

Parameters

Name	Type	Description	Default
agg	str	The type of aggregation to perform. Can be either “att” or “cohort” or “period”. Default is “att”. If “att”, computes the average treatment effect on the treated. If “cohort”, computes the average treatment effect by cohort. If “period”, computes the average treatment effect by period.	`'period'`
weighting	str	The type of weighting to use. Can be either ‘shares’ or ‘variance’.	`'shares'`

Returns

Name	Type	Description
	None

SaturatedEventStudy.summary

summary()

Get summary table.

SaturatedEventStudy.test_treatment_heterogeneity

test_treatment_heterogeneity()

Test for treatment heterogeneity in the event study design.

Parameters

Name	Type	Description	Default
by	str	The type of test to perform. Can be either “cohort” or “time”. Default is “cohort”. If “cohort”, tests for treatment heterogeneity across cohorts as in Lal (2025). See https://arxiv.org/abs/2503.05125 for details.	required

SaturatedEventStudy.tidy

tidy()

Tidy result dataframe.

SaturatedEventStudy.vcov

vcov()

Get the covariance matrix.

Returns

Name	Type	Description
	pd.DataFrame	A DataFrame containing the covariance matrix.