The family of shifted log transformations helps correct nonnormality and skewness. The Automatic Shifted Log transformation of Feng, Hannig, and Marron automatically selects the shift parameter that minimizes either
- the Anderson-Darling statistic (to correct nonnormality)
- the sample skewness (to correct skewness).
from autoshiftedlog import autoshiftedlog
data = numpy.array([1, 1, 2, 2, 2.5, 3, 3, 4.5, 6, 8, 10, 20, 30, 80, 0, numpy.nan])
data = autoshiftedlog(data)