Some colleagues have measured pollutants (PCB, DDT, HCB, etc) in arctic cod and haddock. I was asked to assist with data analysis and plotting for the paper, and decided this was a good opportunity to learn more about the Python libraries. I started out with 'pandas', 'matplotlib', and 'scikit-learn', but ended up using the more full-featured regression from the 'statsmodels' library.
In the end, my program reads the data as CSV files, and outputs plots and a LaTeX source file that produces the supplementary document as a PDF. I'm particularly happy about the code that inspects the output from 'statsmodels' and converts it into sensible units. The rather obscure regression parameters in a log-transformred domain is converted into a percentage-wise decline per year, or a percentage-wise increase in pollutant concentration from doubling fish size.
And the happy news is that as expected (but quite undercommunicated, I think), pollutant levels have decreased substantially in the last decades.
Paper available here, please cite it if you make use of data or code from this repository: http://authors.elsevier.com/sd/article/S0013935119300829