Distance Sampling automation through python and Distance sofware
Python module for AUtomated DIstance SAMpling analyses
This module interfaces distance sampling analysis engines from Distance software, and possibly others in the future ; thus, it has been designed in order to make it easier :
- to run (in parallel) numerous Distance Sampling analyses with many (many) parameter variants on many field observation samples (possibly using some optimisation techniques for automated computation of right and left distance truncations),
- to select the best analysis variant results through a mostly automated process, based on customisable statistical quality indicators,
- to produce partly customisable reports in spreadsheet (numerical results only) and HTML formats (more complete, with full-featured plots like in Distance, and more).
As for now, only the Windows MCDS.exe V6 engine and Point Transect analyses are supported, and so, it runs only under Windows.
The module itself:
- python 3.8+
- pandas 0.25+
- matplotlib 3.1+
- jinja2 2.10+
- zoopt 0.4+
- plotly (sometimes)
You can install pyaudisam from PyPI in your current python environment (conda or venv, whatever):
$ pip install pyaudisam
As a python package, pyaudisam can be used through its python API.
But there's also a command-line interface: try and run it with the -h/--help option.
python -m pyaudisam --help
Whichever method, the best way to go is to read the concrete quick-start guide : see Documentation below.
- a short "how it works" guide to understand the basics (also in French),
- a concrete "quick-start" guide with a real life use case and relevant field data to play with,
- another similar but shorter concrete "quick-start guide" (in French) (command-line only) with the full field data set of the "ACDC 2019" birding study.
Note: You can also get a detailled idea of how to use pyaudisam python API by playing with the fully functional jupyter notebook tests/valtests.ipynb (see below Running tests for how to obtain and run it).
- complete the quick start guides above by other small and focused articles to explain some mandatory details:
- how to build a sample or analysis specification workbook (see a short draft in analyser.py:273),
- write a technical documentation of the whole module,
- write a guide for building the module API documentation (sphinx should work out of the box as reStructured text has been used in docstrings),
some tests are fully automated : after installing pytest, simply run it:
python -m build
Note: Don't care about warnings about pyaudisam.mcds and pyaudisam.report being recognised as importable, but being absent from setuptools' packages configuration ... these folders simply contain pyaudisam config. and data files, no python code at all.
Merge requests are very welcome !
And if you are lacking ideas, here are some good ones ;-)
To do list
- finish tests automation (move tests/unintests.ipynb and tests/valtests.ipynb notebooks code to pytest scripts),
- make pyaudisam work under Linux / Mac OS (all python: OK, but calling MCDS.exe):
- or: through some kind of external client-server interface to MCDS.exe (that runs only under Windows),
- or: by porting MCDS to Linux (closed Fortran source, but old, so might be obtained through a polite request to this Distance Sampling forum ; BUT, needs an IMSL license, which is horribly expensive).
- or: by rewriting MCDS from scratch,
- or: by rewriting MCDS using the MRDS Distance package, meaning some kind of interface to R,
- build a GUI for pyaudisam command-line (with some kind of "project" concept, and parameter set template, and ...),
- add support line transects (only point transects for the moment),
- add support for the co-variates feature of MCDS,
Some formal things that I don't plan to change (let's concentrate on substantive content) :-)
- this code is not blacked or isorted or fully conform to pep8 (but it's clean, commented, and it works),
- the identifier naming scheme used is old-fashioned: camel case everywhere.
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Hashes for pyaudisam-1.0.1-py3-none-any.whl