Benjamin T. Vincent

Polish the Brexit synthetic control example

1

- [ ] Add in more explanation - [ ] Add reference to similar work - [ ] Add caveats - [ ] Make the take home conclusion clearer. Get...

regression discontinuity: allow the treatment to be `>=` or `<=` the threshold

1

At the moment, the assumption is that the units above the threshold are treated. But this absolutely is not always going to be true. So we need to allow for...

Add AR model for interrupted time series

2

#2 added a very simple interrupted time series example with no predictors. But it would be good to add another example where there is more temporal structure. This would then...

Error when `treated` variable dtype is integer

1

Known problem for regression discontinuity, possibly for other experiments... When the `treatment` column data is integer (0/1) then we get an error, it currently only works when the dtype is...

try to get dynamic version in `pyproject.toml`

3

AFTER #105... In theory we should be able to have ONE version number in `causalpy/version.py` and get `pyproject.toml` to extract that. See the instructions here, https://packaging.python.org/en/latest/guides/single-sourcing-package-version/ But when working on...

Improve docs through increased use of intersphinx

1

1. Add more intersphinx mappings in `conf.py`: ``` "pandas": ("https://pandas.pydata.org/docs/", None), "scikit-learn": ("https://scikit-learn.org/stable/", None), ``` And also `arviz` when referring to `idata` for example. 2. Use external intersphinx references when...

Add uncertainty for the traditional (non-Bayesian) models

2

Suggested by @juanitorduz. Would be good to get measures of uncertainty for the non-Bayesian models. Could use: - https://github.com/scikit-learn-contrib/MAPIE - https://github.com/statsmodels/statsmodels ## Statsmodels notes * Using R-style formulas https://www.statsmodels.org/dev/example_formulas.html (also...

People often centre the running variable, such that the threshold is at zero. This will change the parameters which is very relevant for the Bayesian model because of the priors....

Add example for dynamic differences in differences

1

- [ ] In the current bank run example notebook... Clarify in the current example with multiple pre/post observations that this only makes sense if we assume the treatment effect...

At the moment all the Bayesian `ExperimentalDesign` classes have an unruly set of `idata` properties. In an ideal world we would be able to consolidate all these into a single...