For display in findings summaries we can now allow arbitrary strength
of evidence binning. We simply pass in a dict with the strength (as
float) as the key and the string-representation that should appear
in the table as value.
When first importing bibtex entries, if there are e.g. double entries,
the library gives an 'unsuppressable' warning which will always
appear in the manuscript otherwise. This simply ensures no warning
is displayed by turning off logging for the duration of their use.
Since we add the studies that bring findings for the finding tables manually,
this ensures that we spot any mistakes when entering them, or similar
discrepancies on the other side of the raw data.
Instead of it lying in the src directory (where, reasonably, only python
data extraction, processing and modelling code should lie), I turned
the retorque filter into a simple quarto extension.
Could reasonably be repackaged separately from this repo since I believe
other people could profit off it.
Validity calculation belongs to the modelling, so we put it into the
validity module.
Extracting our matrix is a processing step so we made its own matrix
module and put it in their.
Should hopefully provide better separation of concerns going forward.
From strength of findings to the more general validity module, which can then
in turn contain the 'add_to_findings' function which unsurprisingly adds
validities to findings. Makes more sense to me.
Added main findings for structural policies. Moved environmentally-focused
infrastructure projects into structural (infrastructure) policies since
they are not in fact institutional policies.
Some studies were missing crucial fields (prelim studies mostly) which means they
are dropped from being counted in 'by_intervention' aggregation. While
ultimately those will have to have their fields filled correctly, for now we can
simply ensure all studies are aggregated correctly per intervention.
All findings tables can use the validities functionality to add strength of
evidence (internal/external) to themselves. Generalized the function to
work for any main findings csv (to dataframe) table not just institional
findings.
- for weak evidence, + for substantiated evidence, ++ for strong evidence.
Currently separates between internal and external, and uses the bins:
0.0-2.9 for weak
3.0-5.9 for substantiated
6.0-... for strong evidence
Marty Oehme