| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
|
|
Major news:
New cross-sectional models
Beta Regression
BetaModel estimates a regression model for dependent variable in
the unit interval such as fractions and proportions based on the
Beta distribution. The Model is parameterized by mean and precision,
where both can depend on explanatory variables through link functions.
Ordinal Regression
statsmodels.miscmodels.ordinal_model.OrderedModel implements
cumulative link models for ordinal data, based on Logit, Probit or
a userprovided CDF link. Distributions
Copulas
Statsmodels includes now basic support for mainly bivariate copulas.
Currently, 10 copulas are available, Archimedean, elliptical and
asymmetric extreme value copulas. CopulaDistribution combines a
copula with marginal distributions to create multivariate distributions.
Count distribution based on discretization
DiscretizedCount provides count distributions generated by discretizing
continuous distributions available in scipy. The parameters of the
distribution can be estimated by maximum likelihood with
DiscretizedModel.
Bernstein Distribution
BernsteinDistribution creates nonparametric univariate and multivariate
distributions using Bernstein polynomials on a regular grid. This
can be used to smooth histograms or approximate distributions on
the unit hypercube. When the marginal distributions are uniform,
then the BernsteinDistribution is a copula. Statistics
Brunner Munzel rank comparison
Brunner-Munzel test is nonparametric comparison of two samples and
is an extension of Wilcoxon-Mann-Whitney and Fligner-Policello
tests that requires only ordinal information without further
assumption on the distributions of the samples. Statsmodels provides
the Brunner Munzel hypothesis test for stochastic equality in
rank_compare_2indep but also confidence intervals and equivalence
testing (TOST) for the stochastically larger statistic, also known
as Common Language effect size.
Nonparametric
Asymmetric kernels
Asymmetric kernels can nonparametrically estimate density and
cumulative distribution function for random variables that have
limited support, either unit interval or positive or nonnegative
real line. Beta kernels are available for data in the unit interval.
The available kernels for positive data are “gamma”, “gamma2”,
“bs”, “invgamma”, “invgauss”, “lognorm”, “recipinvgauss” and
“weibull” pdf_kernel_asym estimates a kernel density given a
bandwidth parameter. cdf_kernel_asym estimates a kernel cdf. Time
series analysis
Autoregressive Distributed Lag Models
ARDL adds support for specifying and estimating ARDL models, and
UECM support specifying models in error correction form.
ardl_select_order simplifies selecting both AR and DL model orders.
bounds_test implements the bounds test of Peseran, Shin and Smith
(2001) for testing whether there is a levels relationship without
knowing teh orders of integration of the variables.
Fixed parameters in ARIMA estimators
Allow fixing parameters in ARIMA estimator Hannan-Rissanen
(hannan_rissanen) through the new fixed_params argument
|
|
|
|
They now have a tool dependency on py-setuptools instead of a DEPENDS
|
|
|
|
All checksums have been double-checked against existing RMD160 and
SHA512 hashes
|
|
|
|
Many many changes including
Oneway ANOVA-type analysis
~~~~~~~~~~~~~~~~~~~~~~~~~~
Several statistical methods for ANOVA-type analysis of k independent samples
have been added in module :mod:`~statsmodels.stats.oneway`. This includes
standard Anova, Anova for unequal variances (Welch, Brown-Forsythe for mean),
Anova based on trimmed samples (Yuen anova) and equivalence testing using
the method of Wellek.
Anova for equality of variances or dispersion are available for several
transformations. This includes Levene test and Browne-Forsythe test for equal
variances as special cases. It uses the `anova_oneway` function, so unequal
variance and trimming options are also available for tests on variances.
Several functions for effect size measures have been added, that can be used
for reporting or for power and sample size computation.
Multivariate statistics
~~~~~~~~~~~~~~~~~~~~~~~
The new module :mod:`~statsmodels.stats.multivariate` includes one and
two sample tests for multivariate means, Hotelling's t-tests',
:func:`~statsmodels.stats.multivariate.test_mvmean`,
:func:`~statsmodels.stats.multivariate.test_mvmean_2indep` and confidence
intervals for one-sample multivariate mean
:func:`~statsmodels.stats.multivariate.confint_mvmean`
Additionally, hypothesis tests for covariance patterns, and for oneway equality
of covariances are now available in several ``test_cov`` functions.
New exponential smoothing model: ETS (Error, Trend, Seasonal)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
- Class implementing ETS models :class:`~statsmodels.tsa.exponential_smoothing.ets.ETSModel`.
- Includes linear and non-linear exponential smoothing models
- Supports parameter fitting, in-sample prediction and out-of-sample
forecasting, prediction intervals, simulation, and more.
- Based on the innovations state space approach.
Forecasting Methods
~~~~~~~~~~~~~~~~~~~
Two popular methods for forecasting time series, forecasting after
STL decomposition (:class:`~statsmodels.tsa.forecasting.stl.STLForecast`)
and the Theta model
(:class:`~statsmodels.tsa.forecasting.theta.ThetaModel`) have been added.
See 0.12.0-0.12.2 at https://www.statsmodels.org/stable/release/
for the full story, including deprecations.
|
|
Install the new interchangeable BLAS system created by Thomas Orgis,
currently supporting Netlib BLAS/LAPACK, OpenBLAS, cblas, lapacke, and
Apple's Accelerate.framework. This system allows the user to select any
BLAS implementation without modifying packages or using package options, by
setting PKGSRC_BLAS_TYPES in mk.conf. See mk/blas.buildlink3.mk for details.
This commit should not alter behavior of existing packages as the system
defaults to Netlib BLAS/LAPACK, which until now has been the only supported
implementation.
Details:
Add new mk/blas.buildlink3.mk for inclusion in dependent packages
Install compatible Netlib math/blas and math/lapack packages
Update math/blas and math/lapack MAINTAINER approved by adam@
OpenBLAS, cblas, and lapacke will follow in separate commits
Update direct dependents to use mk/blas.buildlink3.mk
Perform recursive revbump
|
|
Major Features:
- Allow fixing parameters in state space models
- Add new version of ARIMA-type estimators (AR, ARIMA, SARIMAX)
- Add STL decomposition for time series
- Functional SIR
- Zivot Andrews test
- Added Oaxaca-Blinder Decomposition
- Add rolling WLS and OLS
- Replacement for AR
Performance Improvements:
- Cythonize innovations algo and filter
- Only perform required predict iterations in state space models
- State space: Improve low memory usability; allow in fit, loglike
|
|
This is a major release from 0.9.0 and includes a number new
statistical models and many bug fixes.
Highlights include:
* Generalized Additive Models. This major feature is experimental and
may change.
* Conditional Models such as ConditionalLogit, which are known as
fixed effect models in Econometrics.
* Dimension Reduction Methods include Sliced Inverse Regression,
Principal Hessian Directions and Sliced Avg. Variance Estimation
* Regression using Quadratic Inference Functions (QIF)
* Gaussian Process Regression
|
|
Remove some .so files from the PLIST that are not built for me
nor in mef's 9.0 bulk build.
|
|
|
|
0.9.0:
The Highlights
--------------
statespace refactoring, Markov Switching Kim smoother
3 Google summer of code (GSOC) projects merged - distributed estimation - VECM and enhancements to VAR (including cointegration test) - new count models: GeneralizedPoisson, zero inflated models
Bayesian mixed GLM
Gaussian Imputation
new multivariate methods: factor analysis, MANOVA, repeated measures within ANOVA
GLM var_weights in addition to freq_weights
Holt-Winters and Exponential Smoothing
|
|
The main features of this release are several new time series models based on the statespace framework, multiple imputation using MICE as well as many other enhancements. The codebase also has been updated to be compatible with recent numpy and pandas releases.
Statsmodels is using now github to store the updated documentation which is available under http://www.statsmodels.org/stable for the last release, and http://www.statsmodels.org/dev/ for the development version.
|
|
Packaged for wip by Kamel Ibn Aziz Derouiche and myself.
Statsmodels is a Python package that provides a complement to scipy
for statistical computations including descriptive statistics and
estimation and inference for statistical models
|
