`bucky.util.power_transforms`#

Simple power transformation classes.

Module Contents#

Classes#

`BoxCox`	Wrapper class for a Box-Cox transformer.
`YeoJohnson`	Wrapper class for a Yeo-Johnson transformer.

Functions#

`boxcox`(y, lam, lam2=None)	Box-Cox tranform, batched in the first dimension.
`fit_lam`(y, yj=False, lam_range=(-2, 2, 0.1))	Fit lambda of a power transform using grid search, taking the the most normally distributed result.
`inv_boxcox`(y, lam1, lam2)	Inverse Box-Cox tranform, batched in the first dimension.
`inv_yeojohnson`(y, lam)	Inverse Yeo-Johnson tranform, batched in the first dimension.
`norm_cdf`(x, mu, sigma)	Normal distribution CDF, batched.
`yeojohnson`(y, lam)	Yeo-Johnson tranform, batched in the first dimension.

Attributes#

bucky.util.power_transforms.EPS = 1e-08#

class bucky.util.power_transforms.BoxCox[source]#

Wrapper class for a Box-Cox transformer.

fit(self, y)[source]#: Fit the batched 1d variables in y, store the lambdas for the inv transform.

inv(self, y)[source]#: Inverse tranform using the fitted lambda values.

class bucky.util.power_transforms.YeoJohnson[source]#

Wrapper class for a Yeo-Johnson transformer.

fit(self, y)[source]#: Fit the batched 1d variables in y, store the lambdas for the inv transform.

inv(self, y)[source]#: Inverse tranform using the fitted lambda values.

bucky.util.power_transforms.boxcox(y, lam, lam2=None)[source]#: Box-Cox tranform, batched in the first dimension.

bucky.util.power_transforms.fit_lam(y, yj=False, lam_range=(- 2, 2, 0.1))[source]#: Fit lambda of a power transform using grid search, taking the the most normally distributed result.

bucky.util.power_transforms.inv_boxcox(y, lam1, lam2)[source]#: Inverse Box-Cox tranform, batched in the first dimension.

bucky.util.power_transforms.inv_yeojohnson(y, lam)[source]#: Inverse Yeo-Johnson tranform, batched in the first dimension.

bucky.util.power_transforms.norm_cdf(x, mu, sigma)[source]#: Normal distribution CDF, batched.

bucky.util.power_transforms.yeojohnson(y, lam)[source]#: Yeo-Johnson tranform, batched in the first dimension.