estimation.rwolf

estimation.rwolf(models, param, reps, seed, sampling_method='wild-bootstrap')

Compute Romano-Wolf adjusted p-values for multiple hypothesis testing.

For each model, it is assumed that tests to adjust are of the form “param = 0”. This function uses the wildboottest() method for running the bootstrap, hence models of type Feiv or Fepois are not supported.

Parameters

Name	Type	Description	Default
models	list[Feols] or FixestMulti	A list of models for which the p-values should be computed, or a FixestMulti object. Models of type Feiv or Fepois are not supported.	required
param	str	The parameter for which the p-values should be computed.	required
reps	int	The number of bootstrap replications.	required
seed	int	The seed for the random number generator.	required
sampling_method	str	Sampling method for computing resampled statistics. Users can choose either bootstrap(‘wild-bootstrap’) or randomization inference(‘ri’)	'wild-bootstrap'

Returns

Type	Description
pd.DataFrame	A DataFrame containing estimation statistics, including the Romano-Wolf adjusted p-values.

Examples

from pyfixest.estimation import feols
from pyfixest.utils import get_data
from pyfixest.multcomp import rwolf

data = get_data().dropna()
fit = feols("Y ~ Y2 + X1 + X2", data=data)
rwolf(fit.to_list(), "X1", reps=9999, seed=123)

fit1 = feols("Y ~ X1", data=data)
fit2 = feols("Y ~ X1 + X2", data=data)
rwolf_df = rwolf([fit1, fit2], "X1", reps=9999, seed=123)
rwolf_df