• A minor fix on retrieving version information with importlib-metadata.

• Drop support for Python 3.5.
• (breaking change) adopy.base.Task takes labels of response variables as the argument responses. Instead, the possible values for the response variables should be given to adopy.base.Engine as an argument named grid_response.
• (breaking change) adopy.base.Model takes a log likelihood function for an argument func now, instead of the probability function for a single binary response variable. The log likelihood function can take multiple response variables.
• (breaking change) The compute method in adopy.base.Model provides the log likelihood function now, instead of the probability of a single binary response variable.
• Using multiple response variables is available now!
• adopy.base.Engine now can update multiple observations, given as a list of designs and a list of corresponding responses into code:design and response arguments, respectively.
• Now, you can choose what datatype to use for adopy.base.Engine, with an argument named dtype. The default is numpy.float32.
• Remove unused types at base (#26; contributed by @NicholasWon47).

A minor version update due to PyPI configuration. It has the exact same features as ADOpy 0.3.0.

The first version that is released publicly.

• General changes
  • Now, we fully updated the README file and documentation.
  • Example codes are provided in the /examples directory.
• Base classes (adopy.base)
  • Possible responses (responses) are now stored in task classes (previously in Engine classes).
  • The engine class gets a new function, reset(), in order to reset an engine object to an initial state.
• Choice under risk and ambiguity task (adopy.tasks.cra)
  • Two design variables, prob and ambig, are renamed to p_var and a_var, respectively.
• Delay discounting task (adopy.tasks.dd)
  • The module name for the delay discounting task is changed to adopy.tasks.dd (previously adopy.tasks.ddt), in order to make it consistent to other tasks (e.g., adopy.tasks.cra). Also, the task and engine classes are renamed to TaskDD and EngineDD (previously TaskDDT and EngineDDT).
  • 4 design variables are renamed to t_ss, t_ll, r_ss and r_ll in order to make it consistent to notations in math equations.
  • In all model classes for the DD task, tau (inverse temperature) is ordered as the last parameter.
  • The model class for Generalized Hyperbolic model, ModelGH, is renamed to ModelHPB for Hyperboloid model.
• Psychometric function estimation (adopy.tasks.psi)
  • The task class for a 2-alternative forced choice task is renamed to Task2AFC (previously TaskPsi). N-AFC task will be supported in future updates (see Issue #18).
  • The model class with a shape of Normal CDF is renamed to ModelProbit (previously ModelNormal).

• Support only for Python 3.5+, excluding Python 2.7.
• Rename adopy.tasks.psi.TaskPsi to adopy.tasks.psi.Task2AFC.
• Add docs powered by Sphinx.

• Fix adopy.functions.get_nearest_grid_index which had made posterior updating malfunction.
• Use autopep8 instead of yapf for develop environment.
• Fix staircase implementation for Psychometric function.

• Fix Engine.post_cov due to the unexpected usage of pandas.DataFrame.

• Use three base classes: Task, Model, and Engine.
• Provide three pre-implemented tasks: psi (Psychometric function), cra (Choice under risk and ambiguity), and ddt (Delayed discounting task).