globaldothealth / monkeypox-report Goto Github PK

Since outbreak_associated is no longer included in Nextstrain TSV download, use clade membership B.1 as the primary indicator of outbreak, as well as include the two A.2 genomes from USA.

Flag outliers in new data.

Create www.monkeypox.global.health/archives.html which will have list of all reports

Consider expanding Figure 4 legend to include the following:

"Figure 4: Age and gender distribution of monkeypox confirmed cases in 2022. X% of confirmed cases include sufficient metadata to be included in this figure. Of these cases, X% include reported age ranges that span multiple age categories. Under such circumstances, counts are evenly distributed among the categories that capture the reported age range. "

I have the lines of code to calculate these percentages.

@Mougk thoughts?

Originally posted by @tvarrelman in globaldothealth/monkeypox#79 (comment)

Can't do it in the usual UI as there's no pull request to reference, so here's the points I observed.

• monkeypox-report/src/test_build.py

  Line 105 in b09edfd

  assert all(url.endswith("csv") for url in build.get_archives_list("csv"))

  consider using monkeypatch to substitute requests.get here, or any problem accessing GitHub will make the test runner exit.

• monkeypox-report/src/build.py

  Line 79 in b09edfd

  logging.error("Failed to get archives list, aborting")

  (and elsewhere this pattern appears) raise an exception and log/exit in if __name__ == '__main__' so that you can also test unhappy paths.

• monkeypox-report/src/test_build.py

  Line 132 in b09edfd

  def test_counts():

  there are a lot of numbers the same in this comparison: how do I know that n_countries_suspected_only is 1 because it correctly counted those, not because it actually used the same logic for n_countries_discarded? My preference would be to introduce functions that calculate the summary statistics separately and test that they work correctly, which might help with:

• monkeypox-report/src/build.py

  Line 166 in b09edfd

  countries_suspected_or_confirmed = set(

  there's a bit of unnecessary complexity making this function hard to read: sometimes sets are calculated as unions/intersections of other sets; sometimes not; sometimes the same set is calculated multiple times; many of them are assigned values in L166-175 but then others are inlined in the return value or even walrus-assigned in the return value and used again on a subsequent line. Maybe refactoring toward helper functions would make it easier to understand what's going on: this is probably the place where errors like including the wrong status in the calculation of a statistic will occur.

• monkeypox-report/src/build.py

  Line 225 in b09edfd

  def mid_bucket_age(age_interval: str) -> float:

  I don't understand why the mid-bucket-age of "50" isn't 50. I'm happy for that to be correct behaviour, but if so could you write a docstring explaining it please? :)

• monkeypox-report/src/build.py

  Line 243 in b09edfd

  51 - 60, 61 - 70, 71 - 80, 81 -

  in G.h we have an upper bound of 120 on age, which is not quite the oldest a human ever gets but acts as a QC check against typos. Worth implementing here?

• monkeypox-report/src/test_build.py

  Line 169 in b09edfd

  [

  the negative (raise ValueError) case isn't tested.

• monkeypox-report/src/build.py

  Line 268 in b09edfd

  return {

  nit: there's a function to calculate occurrences as percentages from data frames that can be extracted here.
• there is a whole lot of setup code in the first 101 lines of test_build.py that seems to mostly generate dictionaries with random content that are used in a two-line test file. Could that be replaced with a constant dictionary, maybe loaded from json or yaml?

• monkeypox-report/src/figures/genomics.py

  Line 30 in b09edfd

  genome_data = genome_data[genome_data.date >= "2022-05"]

  this seems like a big inconsistency: if we don't know which cases are associated with the outbreak then we will consider all cases after May 2022, which includes endemic cases (though maybe not all of them; I don't know if the genomic data is only the West African clade).

I ran the following command

michael@plex:~/code/globalhealth-monkeypox-report$ poetry run python src/build.py bucket

I get the following error, why?

botocore.exceptions.NoCredentialsError: Unable to locate credentials

Figure Legend 1A could be improved. suggestion: a) make smaller, b) put under the Map and horizontal c) Potentially have a zoom for Europe now that there are many lines between European countries.

Fig 1A shows the travel history. At present, there is a lot of travel within Europe which causes overlapping lines. It would be nice to have a zoom out box below the current figure which shows data within Europe only.

Number of countries in text (n = 62) not the same than in figure 1B. Should be consistent. There were some that have discarded cases (perhaps change sentence to: "In X countries suspected or confirmed cases were reported. X confirmed transmission, X continue to have suspected cases only, and X countries discarded cases that were suspected."