clayh53

<script>
/* === routine to wrap img tags in figure tag and follow it with figcaption
This script collects all the image tags on a page. The routine loops through all
the image tags, and creates two new elements - a <figure> and a <figcaption>. 
It then tests if the class is "not-figure", and skips to the next img tag if it 
contains this class. The routine then extracts the class or classes from the image tag
and transfers them to the figure tag, since in Kramdown markdown parsing, an 
image Markdown written like ![image alt name](image_url,"image title"){:.classname 1 .classname 2}
will attach the classes to the image when they are needed in the figure
*/

clayh53

begin = lag+1 # position where predictions start

df['predict_cum'] = df['predicted_death'][lag+1:].cumsum()
df['predict_cum'] = df['predict_cum'] + current_death_total

fig,ax = plt.subplots(figsize=(12,6))
plt.grid(axis='both',linestyle=':',linewidth='1')
sns.set()
sns.set_style("whitegrid")
ax.plot(df.date,df['deathIncrease'].rolling(window=rolling_avg).mean(),color='red',label='14 day moving average new daily infections')

clayh53

df['cumulative_deaths'] = df['deathIncrease'].cumsum(axis=0)
current_death_total = df.at[df.index[-1]+lag+1,'cumulative_deaths'] # picks out the last actual death cumulative total
print(current_death_total)

fig,ax = plt.subplots(figsize=(12,6))
plt.grid(axis='both',linestyle=':',linewidth='1')
sns.set()
sns.set_style("whitegrid")
ax.plot(df.date,df['deathIncrease'].rolling(window=rolling_avg).mean(),color='red',label='14 day moving average new daily infections')
ax.plot(df.date, df['predicted_death'].rolling(window=rolling_avg).mean(), color="blue",linestyle=":",label='14 day moving average new daily deaths')

clayh53

from sklearn.metrics import r2_score
# create columns of 14 day smoothed data
start_R_calc = 192 # index value for Aug 1 2020
end_R_calc = 316 # index for last 'real' data day at time of writing
df['deathIncrease-14daysmoothed'] = df['deathIncrease'].rolling(window=rolling_avg).mean()
df['predicted_Death-14daysmoothed'] = df['predicted_death'].rolling(window=rolling_avg).mean()
model = df[start_R_calc:end_R_calc]['deathIncrease-14daysmoothed']
prediction = df[start_R_calc:end_R_calc]['predicted_Death-14daysmoothed']

R2 = r2_score(model,prediction)

clayh53

df = df.append(pd.DataFrame({'date': pd.date_range(start=df.date.iloc[-1], periods=-lag, freq='D', closed='right')}))

df = df.reset_index() # resets index after what is essentially appending another dataframe

df['avg_death_factor'] = [death_factor] * len(df.index) # copies value to entire column

df['predicted_death'] = df['positiveIncrease'].shift(-lag) * df['avg_death_factor'] # predicts future deaths from lagged initial cases
# Plot it
fig,ax = plt.subplots(figsize=(12,6))
plt.grid(axis='both',linestyle=':',linewidth='1')

clayh53

fig,ax = plt.subplots(figsize=(12,6))
ax.plot(df.date,df['deathIncrease'].rolling(window=rolling_avg).mean(),color='red',label='14 day moving average new daily infections')
sns.set()
sns.set_style("whitegrid")
ax.plot(df.date, df['predicted_death'].shift(-lag).rolling(window=rolling_avg).mean(), color="blue",linestyle=":",label='14 day moving average predicted new daily deaths')
plt.xlim([start_date,end_date]) # moves the prediction forward in time to match the lag as a check
leg = ax.legend(loc='center', frameon=False,bbox_to_anchor=(0.5, -0.17))
leg2=ax2.legend(loc='center', frameon=False,bbox_to_anchor=(0.49, -0.22))

clayh53

# calculate an average death rate from lagged data
avg_death_from_positive_cases = df["death_prediction_scalar"].rolling(window=rolling_avg).mean()
stable_period = 45
start = lag - stable_period # calculates the index from end of dataframe to start the calculation
death_factor = avg_death_from_positive_cases[start:lag].mean() # average of smoothed data
# copy death scalar to an entire column

df['avg_death_factor'] = [death_factor] * len(df.index) # copies value to entire column

df['predicted_death'] = df['positiveIncrease'] * df['avg_death_factor']

clayh53

# plot a zoomed in look at data since June 1
fig,ax = plt.subplots(figsize=(12,5))
plt.xlim([start_date,end_date])
plt.ylim(0,0.05)
sns.set()
sns.set_style("whitegrid")
ax.plot(df.date,df['death_prediction_scalar'],color='red')
ax.plot(df.date, df["death_prediction_scalar"].rolling(window=rolling_avg).mean(), color="blue")

clayh53

# calculate relationship between lagged new daily infections and current daily deaths
df["death_prediction_scalar"] = df["deathIncrease"].shift(lag)/df["positiveIncrease"] # lags deaths back in time by the lag value for this calculation
fig,ax = plt.subplots(figsize=(12,5))
sns.set()
sns.set_style("whitegrid")
ax.plot(df.date,df['death_prediction_scalar'],color='red')
ax.plot(df.date, df["death_prediction_scalar"].rolling(window=rolling_avg).mean(), color="blue")

clayh53

lag = -21 # number of days from current day that prediction factor occurs
fig,ax = plt.subplots(figsize=(12,6))
ax.plot(df.date,df['positiveIncrease'].rolling(window=rolling_avg).mean(),color='red',label='14 day moving average new daily infections')
ax.set_ylabel('Daily new infections')
ax2=ax.twinx()
ax2.plot(df.date, df["deathIncrease"].shift(lag).rolling(window=rolling_avg).mean(), color="blue",label='14 day moving average new daily deaths')
ax2.grid(False) # turn off grid for second Y axis
ax2.set_ylabel('Daily new deaths')
leg = ax.legend(loc='center', frameon=False,bbox_to_anchor=(0.5, -0.10))
leg2=ax2.legend(loc='center', frameon=False,bbox_to_anchor=(0.49, -0.15))