-
Notifications
You must be signed in to change notification settings - Fork 0
/
BlogTruePrevalence.py
210 lines (138 loc) · 6.38 KB
/
BlogTruePrevalence.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
# -*- coding: utf-8 -*-
"""
Work on positivity rate
Comparing positivity rate for new persons, vs. all tests
"""
# path = 'C:/dev/Covid/'
path = './'
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import covid
import urllib
from scipy import signal
import datetime
import os
#%% Get the data
datapath = '.\\data'
csv_file_pop = os.path.join(datapath, 'Population-Data-WI.csv')
# covid.download_pop_data_wi(csv_file_pop)
popdata = covid.read_pop_data_wi(csv_file_pop)
# covid data
widata = covid.read_covid_data_wi('state')
#%% By Test
# manually downloaded file - positives and tests
test_file = "data\\By_Test_Data_data_2020-10-19.csv"
test = pd.read_csv(test_file)
test = test[test['Measure Names']=='Positive tests']
col_rename = {'Day of displaydateonly': 'Date', 'Positives': 'Positives', 'Totals': 'Tests' }
test = test[col_rename.keys()]
test = test.rename(columns=col_rename)
test['Date'] = pd.to_datetime(test['Date'])
test = test.set_index('Date')
# test.plot(y=['Positives','Tests'])
#%% By People
# cases and new people tested
people_file = "data\\By_Person_Data_data_2020-10-19.csv"
people = pd.read_csv(people_file)
people = people[people['Measure Names']=='People tested positive']
col_rename = {'Day of Result Date': 'Date', 'Positive.y': 'Cases', 'Total.Y': 'New people tested' }
people = people[col_rename.keys()]
people = people.rename(columns=col_rename)
people['Date'] = pd.to_datetime(people['Date'])
people = people.set_index('Date')
# people.plot(y=['Cases', 'New people tested'])
#%% Deaths by death date
deaths_file = "data\\Deaths by day_crosstab_2020-10-14.csv"
deaths_raw = pd.read_csv(deaths_file)
# Note: key is to download the file and then re-save it in Excel specifically
# as csv, otherwise it's actually tab delimited and harder to read in in python
deaths = pd.DataFrame(dict(Date=deaths_raw.T.iloc[2:,0],
Deaths=deaths_raw.T.iloc[2:,1],
Prelim=deaths_raw.T.iloc[2:,2]))
# date leaves out the year, so suffix it properly, then convert to datetime
deaths.Date = deaths.Date.astype(str) + '-2020'
deaths.Date = pd.to_datetime(deaths.Date)
deaths = deaths.set_index('Date')
# make sure everything's numeric
deaths.Deaths = pd.to_numeric(deaths.Deaths)
deaths.Prelim = pd.to_numeric(deaths.Prelim)
#%% Compare cases, tests, deaths by date to as-reported
people['Cases Reported'] = widata.set_index('Date').POS_NEW
people['Tests Reported'] = widata.set_index('Date').TEST_NEW
deaths['Reported'] = widata.set_index('Date').DTH_NEW
people.plot(y=['Cases', 'Cases Reported'])
people.plot(y=['New people tested', 'Tests Reported'])
deaths.plot(y=['Deaths', 'Reported'])
#%% Estimate true new cases
people['CaseAvg'] = people.Cases.rolling(window=7, center=False).mean()
test['PosAvg'] = test.Positives.rolling(window=7, center=False).mean()
test['TestAvg'] = test.Tests.rolling(window=7, center=False).mean()
deaths['DeathAvg'] = deaths.Deaths.rolling(window=7, center=False).mean()
test.plot(y=['PosAvg', 'TestAvg'])
est = pd.DataFrame({'cases': people.CaseAvg, 'positives': test.PosAvg, 'tests': test.TestAvg, 'deaths': deaths.DeathAvg})
inf_const = 1/10
est['infections'] = inf_const * est['cases'] * np.sqrt(popdata['WI'] / est['tests'])
est.plot(y=['cases','infections'])
#%% Compare to Youyang Gu's WI estimate
gu_ifr_file = '..\covid19_projections\implied_ifr\IIFR_US_WI.csv'
# read CSV data into a DataFrame, then convert to a Series
gudata = pd.read_csv(gu_ifr_file)
gudata['Date'] = pd.to_datetime(gudata['date']) + datetime.timedelta(days=14)
gudata = gudata.set_index('Date')
est['Gu Estimate'] = gudata['true_inf_est_7day_ma']
est['Bayer Estimate'] = est['infections']
est['Detected x10'] = est['cases'] * 10
est['Detected Cases'] = est['cases']
# back-dated deaths, assume IFR 1%
backdate = 16
ifr = 0.5
name = 'Deaths (IFR ' + str(ifr) + '%, ' + str(backdate) + ' day shift)'
deaths_temp = deaths['DeathAvg'].copy()
deaths_temp.index = deaths_temp.index - datetime.timedelta(days=backdate)
est[name] = deaths_temp / ifr * 100
# est.plot(title='Wisconsin New Infection Estimates', y=['Detected Cases', 'Gu Estimate', 'Bayer Estimate', name])
est.plot(title='Wisconsin New Infection Estimates', y=['cases', 'infections', name])
#%% Cumulative
plt.figure()
plt.plot(people['Cases'].sort_index().cumsum())
plt.plot(est['infections'].cumsum())
plt.plot(datetime.datetime(2020, 6, 30, 0, 0), 93000,'o')
est['Detected cases'] = est['cases']
est['Estimated infections'] = est['infections']
est['Infection / Case Ratio'] = est['infections'] / est['cases']
est.plot(title='Daily True Infection Estimate', y=['Detected Cases', 'Estimated infections'])
est.plot(title='Infection / Case Ratio', y=['Infection / Case Ratio'])
#%% Improve new infection estimate?
# - do a sort of deconvolution for active cases based on duration
# Results in higher estimates in periods when cases are decreasing
# But it's not super dramatic, and it's also really noisy at this point, would
# need lots more smoothing. Not sure it's very promising.
inf_const = 1/10
est['Current Infections'] = est['cases'] * np.sqrt(popdata['WI'] / est['tests'])
est['New Infections 2'] = 0 * est['Current Infections']
duration = 14
# fill NaN with zeroes or the loop below won't work
est = est.fillna(value=0)
for kk in range(len(est['Current Infections'])):
if kk >= duration:
est['New Infections 2'].iloc[kk] = (est['Current Infections'].iloc[kk]
- est['Current Infections'].iloc[kk-1]
+ est['New Infections 2'].iloc[kk-duration])
elif kk > 0:
est['New Infections 2'].iloc[kk] = (est['Current Infections'].iloc[kk]
- est['Current Infections'].iloc[kk-1])
else:
est['New Infections 2'].iloc[kk] = est['Current Infections'].iloc[kk]
est.plot(y=['Detected Cases', 'Estimated infections', 'New Infections 2'], ylim=[0, 5000])
#%% Improve new infection estimate?
# - revised formula to take account of previously detected cases being taken
# out of the testing pool
duration = 14
D = duration
Npop = 5.8e6
Ncases = est['Detected Cases']
Ntests = est['tests']
est['Dedup factor'] = 1 + 1/D * (np.sqrt(Npop/Ntests) - 1)
est['Infections (Dedup)'] = est['Detected Cases'] * est['Dedup factor']
est.plot(y=['Detected Cases', 'Estimated infections', 'Infections (Dedup)'])