moving median

gnuplot_plot

#!/usr/bin/env gnuplot
  


set timefmt '%Y-%m-%d'
set xdata time
set yrange [1<*:]
set format x '%Y-%m-%d'
set xrange ["2020-03-01"<*:]
set datafile sep ','
set logscale y 10
set terminal png size 800,600; set output 'orig.png';
plot 'this.csv' u 1:2 w l title 'orig'~
set terminal png size 1800,1000; set output 'mm.png';
#plot 'this.csv'  u 1:3 w l title 'mm'
plot 'this.csv'  u 1:4 w l title 'mm 60%ile' ,''  u 1:3 w l title 'mm 40%ile', '' u 1:5 w l title "ma"
set terminal png size 1800,1000; set output 'ma.png';
#plot 'this.csv'  u 1:4 w l title 'ma'
set title "Cas de COVID 2019 en France en nouveau cas testés positifs par millions d'habitants (échelle log10) (jul)\nsource https://ourworldindata.org/coronavirus-data-explorer\nNB: Les tests sous évaluent la réalité"
show title
plot 'this.csv'  u 1:4 w l title 'mediane mouvante à 7jours'# , '' u 1:3 w l title "70% ile 7jours"

mov_med.py

#!/usr/bin/env python3
# curl https://raw.githubusercontent.com/owid/covid-19-data/master/public/data/owid-covid-data.csv > info.csv
import csv
res = dict()
from json import dumps
pp = lambda x:print(dumps(x, indent=4))

with open('info.csv', newline='') as csvfile:

    reader = csv.DictReader(csvfile)
    for row in filter(lambda r:r["iso_code"]=="FRA", reader):
        res[row["date"]]=float(row["new_cases_per_million"])

def generate_and_set_moving_function(compute_window, name):
    def filtre_moving(data, win_size=7, **kw):
        window = []
        res=[]
        for d in data:
            window += [d]
            res+= [compute_window(window, **kw)]
            #print(window)
            if len(window) == win_size:
                window.pop(0)
        assert len(data)==len(res)
        return res
    # ugly workaround around python's unability to make you change the function
    # name on the fly, or parametrize (macro) the function name
    filtre_moving.__name__ = name
    globals()[name]=filtre_moving


moving_function(
    lambda window, **kw:sorted(window)[int(kw.get("ratiotile",.5)*len(window))],
    "moving_ratiotile"
)
moving_function(
    lambda window, **kw: sum(window)/(len(window) or 1),
    "moving_average"
)

WS=7
RT=.2

print("date,new_cases_per_million,moving_%sile_new_cases_per_million,moving_%sile_new_cases_per_million, moving_average" %
      (int(RT*100), int((1-RT)*100)))
for l in zip(res.keys(), res.values(),
    moving_average(res.values(), win_size=WS, ratiotile=RT),
    moving_ratiotile(res.values(), win_size=WS, ratiotile=1-RT),
    moving_ratiotile(res.values(), win_size=WS),
    ):
    print(",".join(map(str,l)))