betabinomial

Submodules

Package Contents

Classes

BetaBinomial

Beta-binomial distribution to perform statistical testing on count data.

Functions

pval_adj(pval[, method, alpha])

Multiple testing correction for p-value matrix obtained
class betabinomial.BetaBinomial(alpha=None, beta=None)

Beta-binomial distribution to perform statistical testing on count data.

Parameters:

  • alpha (np.ndarray, optional) – alpha parameter as column vector of beta-binomial. alpha parameter can be learned with infer function. Defaults to None.

  • beta (np.ndarray, optional) – beta parameter as column vector of beta-binomial. beta parameter can be learned with infer function. Defaults to None.

alpha

alpha parameter as column vector of beta-binomial. alpha parameter can be learned with infer function. Defaults to None.

Type:

np.ndarray

beta

beta parameter as column vector of beta-binomial. beta parameter can be learned with infer function. Defaults to None.

Type:

np.ndarray

Examples

Initilize with alpha and beta vector

>>> BetaBinomial(
>>>     alpha=np.array([[1.], [2.], [3.]])
>>>     beta=np.array([[0.5], [0.1], [2]])
>>> )
BetaBinomial[3]

Examples

Initilize with single alpha and beta values

>>> BetaBinomial(
>>>     alpha=np.array([[1.]])
>>>     beta=np.array([[1]])
>>> )
BetaBinomial[1]

Examples

Initilize without alpha and beta

>>> BetaBinomial()
BetaBinomial[]
infer(k, n, theta=0.001, max_iter=1000)

Infer alpha and beta parameters of beta-binomial from k and n counts.

Parameters:

  • k (np.ndarray) – count matrix of observations.

  • n (np.ndarray) – total number of counts events.

  • theta (float, optional) – Error between iterations to stop inference.

  • max_iter – Maximum number of iterations.

_update(k, n, alpha_old, beta_old)
_convergence(alpha_old, alpha, beta_old, beta, theta)
beta_mean()

The mean of beta distrubution = alpha / (alpha+beta)

mean(n)

The expected number of k E[k] = n * alpha / (alpha+beta)

Parameters:

n (np.ndarray) – total number of counts events.

fold_change(k, n)

Fold change between observed k and E[k]

Parameters:

  • k (np.ndarray) – count matrix of observations.

  • n (np.ndarray) – total number of counts events.

log_fc(k, n)

Log-fold change between observed k and E[k]

Parameters:

  • k (np.ndarray) – count matrix of observations.

  • n (np.ndarray) – total number of counts events.

cdf(k, n)

CDF of beta-binomial distribution with given k and n and inferred alpha and beta parameters.

pval(k, n, alternative='two-sided')

Statistical testing with beta-binomial based on given

Parameters:

  • k (np.ndarray) – count matrix of observations.

  • n (np.ndarray) – total number of counts events.

  • alternative – {‘two-sided’, ‘less’, ‘greater’}

z_score(k, n)

z-score based on the k and n and inferred alpha and beta parameters.

Parameters:

  • k (np.ndarray) – count matrix of observations.

  • n (np.ndarray) – total number of counts events.

intra_class_corr()

Intra or inter class corrections.

variance(n)

Variance of beta-binomial distribution.

Parameters:

n (np.ndarray) – total number of counts events.

__repr__()

Return repr(self).

betabinomial.pval_adj(pval, method='fdr_bh', alpha=0.05)

Multiple testing correction for p-value matrix obtained from BetaBinomial.pval

Parameters:

  • pval (np.ndarray) – matrix of p-values.

  • method (str) – Multiple correction method defined based on statsmodels.stats.multitest.multipletests.