nbi-stat 0.8.5

Statistics tools for teaching at NBI

Statistical tools for teaching at NBI

This package extends some of the existing tools in NumPy and SciPy with some useful features designed to make life easier for the students at the Niels Bohr Institute.

Topics

• Reporting scientific results, including proper rounding
• Tabulation of data useful in Jupyter Notebooks
• Visualisation of data in 1 and many dimensions
• Robust calculations of sample means, variances, and covariances, for unweighted and weighted samples. For weighted samples, both frequency and non-frequency weights are supported.
• Histogramming
• Sampling of arbitrary PDFs
• Curve fitting using
  • Linear least squares
  • Non-linear least squares
  • Maximum likelihood estimates
    • Extended
    • Binned
• Representation of fit confidence contours
• Hyppthesis testing
• Confidence intervals
• Template fitting
• Simultaneous fitting over regions (channels)
• Likelihood calculations

Examples of use

This notebook gives examples of use.

Book on Statistics with Python

The book Statistics Overview - With Python lays out much of the theoretical foundation for the tools available.

Some other notes on statistics is available from the same site, including

• Principle Component Analysis as a more robust alternative to boosted decision trees
• Bootstrap and Jackknife and why you should be careful with these estimates
• Coefficent of determination and why you shouldn't use it

Application Programming Interface Documentation

The API is documented.

2019 © Christian Holm Christensen