raphbacher / fronde-py

A Fixed-point algorithm for Robust Null Distribution Estimation

fronde-py

A Fixed-point algorithm for Robust Null Distribution Estimation.

Ref: Celine Meillier et al, GRETSI 2017 https://hal.archives-ouvertes.fr/hal-01563994

This simple and fast method allows to estimate in a robust way the median and the standard deviation under the null hypothesis of data samples that can be contaminated by a source signal. It is a σ-clipping algorithm based on a fixed-point approach.

To install

python setup.py install

To get robust variance and mean estimators of null distribution of a data vector x

from fronde import fronde
import numpy as np

# create null samples
x=np.random.normal(loc=1,scale=0.5,size=1000)

# add outliers/alternative 
x[:100] = 10

# get robust mean and variance
med,var = fronde(x)

You can also try pregenerated data with example.py

python example.py