bqme package¶

Submodules¶

bqme.distributions module¶

class bqme.distributions.Distribution(parameters_dict: Dict[str, bqme.variables.Variable], name: str)¶

Bases: object

Base class for all distribution Key and values of parameters_dict needs the same order as the input to the distribution in stan.

cdf(x: List[float]) → numpy.ndarray¶

code() → Dict[str, str]¶

domain()¶

logcdf(x: List[float]) → numpy.ndarray¶

logpdf(x: List[float]) → numpy.ndarray¶

pdf(x: List[float]) → numpy.ndarray¶

ppf(q: List[float]) → numpy.ndarray¶

class bqme.distributions.Gamma(alpha: float, beta: float, name: str)¶

Bases: bqme.distributions.Distribution

Container for Gamma Distribution

alphafloat: shape of the Gamma
betafloat: rate of the Gamma

name : str

>>> Gamma(1., 1., name='sigma')
Gamma(alpha=1.0, beta=1.0, name="sigma")
>>> Gamma(1.1, 1.0, name='sigma').code()
{'parameter': 'real<lower=0> sigma;', 'prior': 'sigma ~ gamma(1.1, 1.0);'}

domain() → Tuple[float, float]¶

class bqme.distributions.Lognormal(mu: float, sigma: float, name: str)¶

Bases: bqme.distributions.Distribution

Container for Lognormal Distribution

mufloat: log(rv) has loc mu
sigmafloat: log(rv) has scale sigma

name : str

>>> Lognormal(1., 1., name='sigma')
Lognormal(mu=1.0, sigma=1.0, name="sigma")
>>> Lognormal(1.1, 1.0, name='sigma').code()
{'parameter': 'real<lower=0> sigma;', 'prior': 'sigma ~ lognormal(1.1, 1.0);'}

domain() → Tuple[float, float]¶

class bqme.distributions.Normal(mu: float, sigma: float, name: str)¶

Bases: bqme.distributions.Distribution

Container for Normal Distribution

mufloat: location of the normal
sigmafloat: standard deviation of the normal

name : str

>>> Normal(0, 1, name='mu')
Normal(mu=0, sigma=1, name="mu")
>>> Normal(0, 1, name='mu').code()
{'parameter': 'real mu;', 'prior': 'mu ~ normal(0, 1);'}

domain() → Tuple[float, float]¶

class bqme.distributions.Weibull(alpha: float, sigma: float, name: str)¶

Bases: bqme.distributions.Distribution

Container for Weibull Distribution

alphafloat: shape of the weibull
sigmafloat: scale of the weibull - equivalent to 1/rate

name : str

>>> Weibull(1., 1., name='sigma')
Weibull(alpha=1.0, sigma=1.0, name="sigma")
>>> Weibull(1.1, 1.0, name='sigma').code()
{'parameter': 'real<lower=0> sigma;', 'prior': 'sigma ~ weibull(1.1, 1.0);'}

domain() → Tuple[float, float]¶

bqme.fit_object module¶

class bqme.fit_object.FitObject¶

Bases: object

Base class for the fit object

cdf(x: float, method='mean')¶

Calculates the cdf of x using posterior samples or MAP estimate

xfloat or List[float]: points where the cdf should be evaluated
method: str, default: ‘mean’: if sampling is used possible values are (‘mean’, ‘median’, ‘full’). Return values is the mean over all samples if ‘mean’ is selected. Otherwise median or the full matrix is returned if optimizing is used return values are the pdf evaluation of the model at x.

ret : ndarray

pdf(x: float, method='mean')¶

Calculates the pdf of x using posterior samples or MAP estimate

xfloat or List[float]: points where the pdf should be evaluated
method: str, default: ‘mean’: if sampling is used possible values are (‘mean’, ‘median’, ‘full’). Return values is the mean over all samples if ‘mean’ is selected. Otherwise median or the full matrix is returned if optimizing is used return values are the pdf evaluation of the model at x.

ret : ndarray

ppf(q: float, method='full')¶

Calculates the percent point funtion (ppf) of x using posterior samples or MAP estimate

qfloat or List[float]: points where the ppf should be evaluated. Must be in range (0, 1)
method: str, default: ‘mean’: if sampling is used possible values are (‘mean’, ‘median’, ‘full’). Return values is the mean over all samples if ‘mean’ is selected. Otherwise median or the full matrix is returned if optimizing is used return values are the pdf evaluation of the model at x.

ret : ndarray

class bqme.fit_object.FitObjectOptimizing(model: QM, opt_parameters: Dict)¶

Bases: bqme.fit_object.FitObject

Fit object using MAP estimate of the model

class bqme.fit_object.FitObjectSampling(model: QM, stan_fit_object: StanFit4Model)¶

Bases: bqme.fit_object.FitObject

Fit object using posterior samples of the model. This is an extension of the ‘StanFit4Model’-type by composition.

bqme.models module¶

class bqme.models.GammaQM(alpha: bqme.distributions.Distribution, beta: bqme.distributions.Distribution)¶

Bases: bqme.models.QM

Quantile matching using Gamma distribution

alphaDistribution: Also called the shape of the Gamma
betaDistribution: Also called the rate of the Gamma

>>> from bqme.distributions import Gamma
>>> model = GammaQM(Gamma(1., 1. ,'alpha'), Gamma(1., 1., 'beta'))
>>> model
GammaQM(Gamma(alpha=1.0, beta=1.0, name="alpha"), Gamma(alpha=1.0, beta=1.0, name="beta"))
>>> code = model.code

domain() → Tuple[float, float]¶

class bqme.models.LognormalQM(mu: bqme.distributions.Distribution, sigma: bqme.distributions.Distribution)¶

Bases: bqme.models.QM

Quantile matching using Lognormal distribution

muDistribution: location of the corresponding Normal distribution
sigmaDistribution: scale of the corresponding Normal distribution

>>> from bqme.distributions import Lognormal, Normal
>>> model = LognormalQM(Normal(1., 1. ,'mu'), Lognormal(1., 1., 'sigma'))
>>> model
LognormalQM(Normal(mu=1.0, sigma=1.0, name="mu"), Lognormal(mu=1.0, sigma=1.0, name="sigma"))
>>> code = model.code

domain() → Tuple[float, float]¶

class bqme.models.NormalQM(mu: bqme.distributions.Distribution, sigma: bqme.distributions.Distribution)¶

Bases: bqme.models.QM

Quantile matching using Normal distribution

muDistribution: location of the Normal
sigmaDistribution: scale of the Normal

>>> from bqme.distributions import Normal, Gamma
>>> model = NormalQM(Normal(0., 1., 'mu'), Gamma(1., 1., 'sigma'))
>>> model
NormalQM(Normal(mu=0.0, sigma=1.0, name="mu"), Gamma(alpha=1.0, beta=1.0, name="sigma"))
>>> code = model.code

domain() → Tuple[float, float]¶

class bqme.models.QM(parameters_dict: Dict[str, bqme.distributions.Distribution])¶

Bases: object

Base class for quantile matching (QM) models

parameters_dictDict: Keys are internal names for the priors of the model e.g. ‘mu’, ‘sigma’ for a GaussianQM. Values are the user defined Distributions. Note key must not be identical to value.name.

property code¶: returns the final stan code

compile()¶

domain()¶

optimizing(N: int, q: Tuple[float, …], X: Tuple[float, …]) → StanFit4Model¶

sampling(N: int, q: Tuple[float, …], X: Tuple[float, …]) → StanFit4Model¶

class bqme.models.WeibullQM(alpha: bqme.distributions.Distribution, sigma: bqme.distributions.Distribution)¶

Bases: bqme.models.QM

Quantile matching using Weibull distribution

alphaDistribution: Also called the shape of the Weibull
sigmaDistribution: Also called the scale of the Weibull

>>> from bqme.distributions import Weibull
>>> model = WeibullQM(Weibull(1., 1. ,'alpha'), Weibull(1., 1., 'sigma'))
>>> model
WeibullQM(Weibull(alpha=1.0, sigma=1.0, name="alpha"), Weibull(alpha=1.0, sigma=1.0, name="sigma"))
>>> code = model.code

domain() → Tuple[float, float]¶

bqme.variables module¶

class bqme.variables.ContinuousVariable(value: float, name: str)¶

Bases: bqme.variables.Variable

container for continuous variables (-inf, inf)

class bqme.variables.PositiveContinuousVariable(value: float, name: str)¶

Bases: bqme.variables.Variable

container for positive continous variables (0, inf)

class bqme.variables.Variable(value: float, name: str, lower: float, upper: float)¶

Bases: object

Base class for different kind of variables

approve(value: float) → float¶

bqme package¶

Submodules¶

bqme.distributions module¶

bqme.fit_object module¶

bqme.models module¶

bqme.variables module¶

Module contents¶