Fix pagerank percentile

[jcushman]

I noticed this when a user asked me to explain the meaning of the numbers in our pagerank export, and figured I may as well throw in a fix --

The code to generate pagerank percentiles was obviously intended to give all cases with the same pagerank the same percentile score, but missing a last_score = score line. The upshot is that row one gets a percentile of 1/n, row 2 gets a percentile of 2/n, etc, instead of them all having the same percentile if they have the same score:

id,raw_score,percentile
4,4.076306636737454e-08,0.0
5,4.076306636737454e-08,1.9539073261755196e-07
7,4.076306636737454e-08,3.907814652351039e-07
13,4.076306636737454e-08,5.861721978526558e-07
17,4.076306636737454e-08,7.815629304702078e-07

This just puts in the missing assignment so those would all have percentile 0.0. Note this has the largest effect on the lowest-ranked cases -- about a third of our cases have the same lowest pagerank, so case 1.7 million has a percentile of 33% when it should be 0%. After that there's more gradation so there's less effect from this bug.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (09f8625) 61.97% compared to head (ff5b10a) 61.98%.
Report is 1 commits behind head on develop.

Additional details and impacted files

@@           Coverage Diff            @@
##           develop    #2182   +/-   ##
========================================
  Coverage    61.97%   61.98%           
========================================
  Files          107      107           
  Lines        11817    11818    +1     
========================================
+ Hits          7324     7325    +1     
  Misses        4493     4493           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[bensteinberg]

👍

When this is deployed, I think I'll run fab calculate_pagerank_scores ; fab load_pagerank_scores?

[jcushman]

I think it would be fab export_citation_graph?