did.estimation.lpdid

did.estimation.lpdid(
    data
    yname
    idname
    tname
    gname
    vcov=None
    pre_window=None
    post_window=None
    never_treated=0
    att=True
    xfml=None
)

Local projections approach to estimation.

Estimate a Difference-in-Differences / Event Study Model via the Local Projections Approach.

Parameters

Name	Type	Description	Default
data	DataFrame	The DataFrame containing all variables.	required
yname	str	The name of the dependent variable.	required
idname	str	The name of the id variable.	required
tname	str	Variable name for calendar period.	required
gname	str	Unit-specific time of initial treatment.	required
vcov	(str, dict)	The type of inference to employ. Defaults to {“CRV1”: idname}. Options include “iid”, “hetero”, or a dictionary like {“CRV1”: idname}.	None
pre_window	int	The number of periods before the treatment to include in the estimation. Default is the minimum relative year in the data.	None
post_window	int	The number of periods after the treatment to include in the estimation. Default is the maximum relative year in the data.	None
never_treated	int	Value in gname indicating units never treated. Default is 0.	0
att	bool	If True, estimates the pooled average treatment effect on the treated (ATT). Default is False.	True
xfml	str	Formula for the covariates. Not yet supported.	None

Returns

Name	Type	Description
	DataFrame	A DataFrame with the estimated coefficients.

Examples

import pandas as pd
import pyfixest as pf

url = "https://raw.githubusercontent.com/py-econometrics/pyfixest/master/pyfixest/did/data/df_het.csv"
df_het = pd.read_csv(url)

fit = pf.lpdid(
    df_het,
    yname="dep_var",
    idname="unit",
    tname="year",
    gname="g",
    vcov={"CRV1": "state"},
    pre_window=-20,
    post_window=20,
    att=False
)

fit.tidy().head()
fit.iplot(figsize= [1200, 400], coord_flip=False).show()

To get the ATT, set att=True:

fit = pf.lpdid(
    df_het,
    yname="dep_var",
    idname="unit",
    tname="year",
    gname="g",
    vcov={"CRV1": "state"},
    pre_window=-20,
    post_window=20,
    att=True
)
fit.tidy()

	Estimate	Std. Error	t value	Pr(>|t|)	2.5%	97.5%	N
treat_diff	2.506746	0.071357	35.129648	0.0	2.362413	2.65108	5716.0