CovidData.py

#!/usr/bin/env python
# coding: utf-8

# In[1]:


import numpy as np
import matplotlib.pyplot as plt
import matplotlib.dates as mdates
import pandas as pd
from datetime import datetime
from datetime import timedelta
from dateutil.tz import tzlocal
#get_ipython().magic('matplotlib inline')


# In[2]:


world_poppercap = 7700000000 / 1000000
us_poppercap = 329877505 / 1000000
la_poppercap = 4648794 / 1000000

ticks = 10  #number of X ticks


# In[3]:


def weekofyear(date):
    """
    Convert date to week of year for plotting weekly data
    Intended use for pandas.dataframe.map
    Need to be done before dropping yeah from date
    """
    dt = datetime.strptime(date, '%Y-%m-%d')
    week = int(dt.strftime('%U'))
    
    #adjust for 2021
    if dt.year > 2020:
        week += 52
    
    return week


def rollingavg(y, days = 7):
    y = np.array(y)
    
    output = np.zeros((y.shape[0] * days), dtype=np.float64).reshape((y.shape[0],days))
        
    for idx in range(days,y.shape[0]):
        output[idx] = y[idx-days:idx]
    
    return np.nanmean(output,axis=1)
    
def newstats(df):
    """
        Changes daily totals to new cases and new deaths and adds them to dateframe
    """
    #changing daily totals to new cases
    newcases = np.zeros(df.shape[0], dtype=np.int32)
    newdeaths = np.zeros(df.shape[0], dtype=np.int32)
    deathscolumn = df.columns.get_loc('deaths')
    casescolumn = df.columns.get_loc('cases')

    newcases[0] = df.iloc[0,casescolumn]
    newdeaths[0] = df.iloc[0,deathscolumn]


    for idx in df.index[1:]:
        newcases[idx] = df.iloc[idx,casescolumn] - df.iloc[idx-1,casescolumn]
        newdeaths[idx] = df.iloc[idx,deathscolumn] - df.iloc[idx-1,deathscolumn]

    df['NewCases'] = newcases
    df['NewDeaths'] = newdeaths
    df['NewCasesRollAvg'] = rollingavg(newcases)
    df['NewDeathsRollAvg'] = rollingavg(newdeaths)

def drawindexofCDCdataloss(df,ax, color="magenta", label="CDC Loses Ctrl"):
    pass
    
def getyTicks(df, columns, yticks=8):
    """
        Gets yTicks and returns labels
    """
    dfmax = np.max(df[columns])
    items_per_tick = dfmax / yticks
    out_yticks = np.arange(0,dfmax,items_per_tick)
    #axs[0].set_yticks(la_yticks)
    out_ticklabels = []
    for i in out_yticks:
        if i > 1000000:
            out_ticklabels.append("{0:,.3g}M".format(i/1000000.))
        elif i > 10000:
            out_ticklabels.append("{0:,.3g}K".format(i/1000.0))
        else:
            out_ticklabels.append("{0:,.0f}".format(i))

    return (out_yticks, out_ticklabels)
    
#This is to mark the dataset problem mentioned above

def LACFline(df,ax):
    """
        Adds a line for data adjustment to LA data on 6/19 mentioned above.  nytimes issues 377
        This has become less relavent and so isn't being used
    """
    
    idx = df[df['date']=='06-19'].index.values[0]
    line = ax.axvline(idx,ymin=0,ymax=1,color='r',label="LA Data Adjust", zorder=-100)
    
    return line

def holidays(df,ax):
    """
        Adds lines for major holiday
        
    """
    holidays_list = [
        ('2020-03-25','Memorial Day'),
        ('2020-07-04', '4th of July'),
        ('2020-04-12','Easter'),
        ('2020-09-07','Labor Day'),
        ('2020-11-27','Thanksgiving'),
        ('2021-01-01','New Years')
    ]
    
    label = "Holiday"
    for h in holidays_list:
        #circling back to this
        #line = ax.axvline(df[df['date']==h[0]].index.values[0],ymin=0,ymax=1,color='pink',label=label, zorder=-100)
        #label = None
        pass
    
    #return line
    return None
    
def ladhdatamax(ax):
    """
        Adds a line where the LaDH data stops
    """
    return ax.axvline(orleans_df.index[-1]+1, color='red', label='End LaDH dataset', alpha=0.5)


# ![image.png](attachment:image.png)
# 
# There was an issue with one of the days in louisiana which caused a negative result in the newcases for 6/19
# 
# https://github.com/nytimes/covid-19-data/issues/377

# In[4]:


#state_df = pd.read_csv('./us-states.csv')
state_df = pd.read_csv('./covid-19-data/us-states.csv')  #moved nytimes data to submodule

la_df = state_df[state_df['state'] == 'Louisiana']
la_df = la_df.sort_values(['date'], ascending=True)

#week of year
la_df['week'] = la_df['date'].map(weekofyear)  #maybe this will be used again later

#little bit of a year 2000 problem here
#shorten date  
#la_df['date'] = la_df['date'].map(lambda x: x[5:])

la_df['cases_ravg'] = rollingavg(la_df['cases'])
la_df['deaths_ravg'] = rollingavg(la_df['deaths'])

la_df.index = np.arange(la_df.shape[0])
newstats(la_df)

la_df['NewCasesPerCapita'] = la_df['NewCases'] / la_poppercap
la_df['NewDeathsPerCapita'] = la_df['NewDeaths'] / la_poppercap

la_df['NewCasesPerCapita_ravg'] = rollingavg(la_df['NewCasesPerCapita'])
la_df['NewDeathsPerCapita_ravg'] = rollingavg(la_df['NewDeathsPerCapita'])

la_items_pertick = la_df.shape[0] / ticks
la_ticks = np.append(np.arange(0,la_df.shape[0], la_items_pertick), [la_df.shape[0]-1])
la_items_pertick += 1


la_df.tail(2)


# In[5]:


us_df = pd.read_csv('./covid-19-data/us.csv')  
us_df = us_df.sort_values('date')
us_df[us_df['date'] == '2020-03-09']

#slice to match up with Louisiana data for better comparison
us_df = us_df.iloc[48:,:]

#week of year
us_df['week'] = us_df['date'].map(weekofyear)  #maybe this will be used again later

#little bit of a year 2000 problem ghere
#shorten date
#us_df['date'] = us_df['date'].map(lambda x: x[5:])

us_df['cases_ravg'] = rollingavg(us_df['cases'])
us_df['deaths_ravg'] = rollingavg(us_df['deaths'])

us_df.index = np.arange(us_df.shape[0])
newstats(us_df)

us_df['NewCasesPerCapita'] = us_df['NewCases'] / us_poppercap
us_df['NewDeathsPerCapita'] = us_df['NewDeaths'] / us_poppercap

us_df['NewCasesPerCapita_ravg'] = rollingavg(us_df['NewCasesPerCapita'])
us_df['NewDeathsPerCapita_ravg'] = rollingavg(us_df['NewDeathsPerCapita'])

us_items_pertick = us_df.shape[0] / ticks
us_ticks = np.append(np.arange(0,us_df.shape[0], us_items_pertick),[us_df.shape[0]-1])
us_items_pertick += 1

#used to set xlim to lign up graphs
xdatalen = us_df.index.shape[0]
us_df.tail(2)


# In[6]:


columns = ['date','new_cases','new_deaths','total_cases','total_deaths']
newcolumns = ['date','NewCases','NewDeaths','cases','deaths']  #set the columns to match the other dataframes

world_df = pd.read_csv('./owid-coviddata/public/data/owid-covid-data.csv')
world_df = world_df[world_df['iso_code'] == 'OWID_WRL'][columns]
world_df.columns = newcolumns

world_df = world_df.sort_values('date')
world_df.index = np.arange(world_df.shape[0])
world_df = world_df.iloc[world_df.index[world_df['date'] == '2020-03-09'][0]:,:]

world_df['week'] = world_df['date'].map(weekofyear)
#world_df['date'] = world_df['date'].map(lambda x: x[5:])

world_df['NewCasesRollAvg'] = rollingavg(world_df['NewCases'])
world_df['NewDeathsRollAvg'] = rollingavg(world_df['NewDeaths'])

world_df['cases_ravg'] = rollingavg(world_df['cases'])
world_df['deaths_ravg'] = rollingavg(world_df['deaths'])

world_df.index = np.arange(world_df.shape[0])

world_df['NewCasesPerCapita'] = world_df['NewCases'] / world_poppercap
world_df['NewDeathsPerCapita'] = world_df['NewDeaths'] / world_poppercap

world_df['NewCasesPerCapita_ravg'] = rollingavg(world_df['NewCasesPerCapita'])
world_df['NewDeathsPerCapita_ravg'] = rollingavg(world_df['NewDeathsPerCapita'])

world_items_pertick = world_df.shape[0] / ticks
world_items_per_tick = world_df.shape[0] / ticks
world_ticks = np.append(np.arange(0,world_df.shape[0], world_items_pertick),[world_df.shape[0]-1])


world_df.tail(2)


# In[7]:


df = pd.read_excel("./LaDeptHealth/LA_COVID_TESTBYDAY_PARISH_PUBLICUSE.xlsx")
df.head()


# In[8]:


parishes = sorted(df['Parish'].unique())
print(parishes)


# In[9]:


mindate = df['Lab Collection Date'].min()
print(mindate)


# In[10]:


maxdate = df['Lab Collection Date'].max()
print(maxdate)


# In[11]:


columns = []
dates_raw = np.sort(df['Lab Collection Date'].unique())

dates = sorted(np.datetime_as_string( dates_raw, unit='D'))
for date in dates_raw:
    columns.append(date)

rows_cases = []
rows_tests = []


for parish in parishes:
    parish_mask = df['Parish'] == parish
    
    casecount = []
    testcount = []

    for date in dates_raw:
        
        data = df[np.logical_and( parish_mask,  df['Lab Collection Date']==date)]
        casecount.append(data['Daily Case Count'].iloc[0])
        testcount.append(data['Daily Test Count'].iloc[0])
    
    
    
    assert len(casecount) == len(columns), "Outdata_cases len=%d, but columns len=%d" % (len(outdata),len(columns))
    assert len(testcount) == len(columns), "Outdata_tests len=%d, but columns len=%d" % (len(outdata),len(columns))
    rows_cases.append(casecount)
    rows_tests.append(testcount)


#rows_cases.append(outdata_cases)
#rows_tests.append(outdata_tests)
    
df_parish_cases = pd.DataFrame(rows_cases, columns=columns, index=parishes)
df_parish_tests = pd.DataFrame(rows_tests, columns=columns, index=parishes)
df_parish_cases.loc['Louisiana'] = df_parish_cases.sum()
df_parish_tests.loc['Louisiana'] = df_parish_tests.sum()


    


# In[12]:


df_parish_cases.tail(2)


# In[13]:


df_parish_tests.tail(2)


# In[14]:


df_population = pd.read_csv("./uscensus/co-est2019-alldata.csv" ,encoding='latin1')
df_population = df_population[df_population['STNAME'] == 'Louisiana']
df_population.index = df_population['CTYNAME'].str.replace(" Parish","")
orleans_population = df_population.loc['Orleans']['POPESTIMATE2019']
la_population = df_population.loc['Louisiana']['POPESTIMATE2019']
ebr_population = df_population.loc['East Baton Rouge']['POPESTIMATE2019']


# In[15]:


def processparishdf(parish):
    populationpercap = df_population.loc[parish]['POPESTIMATE2019'] / 1000000
    
    df = pd.DataFrame( df_parish_cases.columns, columns=['date'])
    df['date'] = df['date'].map(lambda x: "%4d-%.2d-%.2d" % (x.year,x.month,x.day))
    df['week'] = df['date'].map(weekofyear)
    #df['date'] = df['date'].map(lambda x: x[5:])
    
    df['NewCases'] = df_parish_cases.loc[parish].values
    df['NewCasesRollAvg'] = rollingavg(df['NewCases'])
    
    cases = []
    cases_num = 0
    for row in df.iterrows():
        cases_num += row[1][2]
        cases.append(cases_num)
    
    df['cases'] = cases
    df['cases_ravg'] = rollingavg(df['cases'])
    
    df['NewCasesPerCapita'] = df['NewCases'] / populationpercap
    df['NewCasesPerCapita_ravg'] = rollingavg(df['NewCasesPerCapita'])
    df = df.iloc[8:,:]
    
    df.index = np.arange(df.shape[0])
    
    items_pertick = df.shape[0] / ticks
    out_ticks = np.append(np.arange(0,df.shape[0], items_pertick), [df.shape[0]-1])
    items_pertick += 1
    
    return (df, out_ticks)
    
(orleans_df, orleans_ticks) = processparishdf('Orleans')
(ebr_df, ebr_ticks) = processparishdf('East Baton Rouge')
(la2_df,la2_ticks) = processparishdf('Louisiana')
(tamm_df,tamm_ticks) = processparishdf('St. Tammany')
    
orleans_df.tail(2)


# In[16]:


#Figure 9

fig,axs = plt.subplots(figsize=(15,10))

#axs.plot(la_df['date'],la_df['NewCases'], label='NyTimes Data')
#axs.plot(la2_df['date'],la2_df['NewCases'], label='LaDH Data')

axs.plot(la_df['date'],la_df['NewCasesRollAvg'], label='NyTimes Data ravg')
axs.plot(la2_df['date'],la2_df['NewCasesRollAvg'], label='LaDH Data ravg')

idx = 0
for label in axs.xaxis.get_ticklabels():
    
    if idx % 10 != 0:
        label.set_visible(False)
        
    idx = idx+1

axs.tick_params(axis='x', rotation=90)
axs.set_title("NyTimes vs La Dept of Health New Case data from Louisiana Rolling Avg (for comparing data sources)")

ladhdatamax(axs)
axs.legend()
fig.savefig("fig9.jpg")


# In[17]:


#Figure 1
fig, axs = plt.subplots(6, figsize=(15,19))

def plotme(df, ax,xticks,addeaths=True):
    


    ax.bar(df.index,df['NewCases'], color='y', label='New Cases', width=1.0)
    ax.plot(df.index,df['NewCasesRollAvg'], color = 'b',label='New Cases 7DRollAvg')
    
    if addeaths:
        ax.bar(df.index,df['NewDeaths'], color='r', label='New Deaths', width=1.0)
        ax.plot(df.index,df['NewDeathsRollAvg'], color = 'g', label='New Deaths 7DRollAvg')

    ax.set_xticks(us_ticks)
    ax.set_xticklabels(us_df.iloc[us_ticks,0])
    xmarg,_ = ax.margins()
    ax.set_xlim(xmax=xdatalen + (xdatalen * xmarg))
    
    drawindexofCDCdataloss(df,ax)
    holidays(df,ax)
    ladhdatamax(ax)    
    yticks,yticklabels = getyTicks(df,'NewCases')
    ax.set_yticks(yticks)        
    ax.set_yticklabels(yticklabels)
    ax.set_ylabel("New Cases")
    
    
    
axs[0].title.set_text('Orleans')
plotme(orleans_df,axs[0],orleans_ticks, addeaths=False)

axs[1].title.set_text("East Baton Rouge")
plotme(ebr_df,axs[1],ebr_ticks,addeaths=False)

axs[2].title.set_text("St. Tammany")
plotme(tamm_df,axs[2],tamm_ticks,addeaths=False)

axs[3].title.set_text('Louisiana (nytimes data)')
#adjustments for LA data issue 4/19
#LACFline(la_df,axs[3])
axs[3].plot(la2_df.index,la2_df['NewCasesRollAvg'], linestyle='--',color = 'b',label='New Cases 7DRollAvg (LaDH)')
plotme(la_df,axs[3],la_ticks, addeaths=False)

axs[4].title.set_text('US')
plotme(us_df,axs[4],us_ticks, addeaths=False)

axs[5].title.set_text('World')
plotme(world_df,axs[5],world_ticks, addeaths=False)

#add lines for summer highs
summstart = pd.to_datetime('2020-05-01')
summend = pd.to_datetime('2020-09-01')
for curr_df, ax in [(orleans_df, axs[0]),(ebr_df,axs[1]),(tamm_df,axs[2]),(la_df,axs[3]),(us_df,axs[4])]:
    dts = pd.to_datetime(curr_df['date'])
    summermask = (dts > summstart) & ( dts < summend)
    summerhigh = curr_df['NewCasesRollAvg'][summermask].max()
    ax.axhline(summerhigh, label="Aprox. Summer high (%.0d)" % summerhigh, color='orange', alpha=0.4)
 
for ax in fig.axes:
    ax.legend()
    
fig.suptitle("New Daily Cases/Deaths Covid-19 La vs US")
fig.tight_layout()
plt.savefig("fig1.jpg")


# In[18]:


#Figure 2
fig, axs = plt.subplots(6, figsize=(15,19))

def plotme(df,ax, xticks, addeaths = True):
    ax.bar(df.index,df['cases'], color='y', label='Cases', width=1.0)
    ax.plot(df.index,df['cases_ravg'], color = 'b',label='Cases RollAvg')
    
    if addeaths:
        ax.bar(df.index,df['deaths'], color='r', label='Deaths', width=1.0)
        ax.plot(df.index,df['deaths_ravg'], color = 'g', label='Deaths RollAvg')
        
    ax.set_xticks(us_ticks)
    ax.set_xticklabels(us_df.iloc[us_ticks,0])
    xmarg,_ = ax.margins()
    ax.set_xlim(xmax=xdatalen + (xdatalen * xmarg))
    
    drawindexofCDCdataloss(df,ax)
    holidays(df,ax)
    ladhdatamax(ax)
    ax.legend(ncol=2)
    

    yticks,yticklabels = getyTicks(df,'cases')
    ax.set_yticks(yticks)        
    ax.set_yticklabels(yticklabels)
    ax.set_ylabel("Cases/Deaths")

axs[0].title.set_text("Orleans")
plotme(orleans_df,axs[0],orleans_ticks, addeaths=False)

axs[1].title.set_text("East Baton Rouge")
plotme(ebr_df,axs[1],ebr_ticks, addeaths=False)

axs[2].title.set_text("St. Tammany")
plotme(tamm_df,axs[2],tamm_ticks,addeaths=False)

axs[3].title.set_text('Louisiana (nytimes data)')
axs[3].plot(la2_df.index,la2_df['cases_ravg'], linestyle='--', color = 'b',label='Cases RollAvg (LaDH)')
plotme(la_df,axs[3],la_ticks,addeaths=True)
#adjustments for LA data issue 4/19
#LACFline(la_df,axs[3])

axs[4].title.set_text('US')
plotme(us_df,axs[4],us_ticks)

axs[5].title.set_text("World")
plotme(world_df,axs[5],world_ticks)

fig.suptitle("Covid-19 Orleans/EBR/La/US vs World Totals")
fig.tight_layout()
plt.savefig("fig2.jpg")


# In[19]:


#Figure 3

fig, axs = plt.subplots(3, figsize=(15,10))

def plotme(df,ax,xticks):

    ax.bar(df.index,df['NewDeaths'], color = 'y', label="New Deaths", width=1.0)
    ax.plot(df.index,df['NewDeathsRollAvg'], color='g', label="New Deaths 7DRollAvg")
    
    ax.set_xticks(us_ticks)
    ax.set_xticklabels(us_df.iloc[us_ticks,0])
    xmarg,_ = ax.margins()
    ax.set_xlim(xmax=xdatalen + (xdatalen * xmarg))
    
    drawindexofCDCdataloss(df,ax)
    holidays(df,ax)
    #ladhdatamax(ax)  #no ladh data on this one
    ax.legend()

    yticks,yticklabels = getyTicks(df,'NewDeaths')
    ax.set_yticks(yticks)        
    ax.set_yticklabels(yticklabels)
    ax.set_ylabel("New Deaths")

axs[0].title.set_text('Louisiana Deaths (nytimes)')

#adjustments for LA data issue 4/19
#LACFline(la_df,axs[0])

plotme(la_df,axs[0],la_ticks)


axs[1].title.set_text('US Deaths')
plotme(us_df,axs[1],us_ticks)

axs[2].title.set_text('World Deaths')
plotme(world_df,axs[2],world_ticks)

#add lines for summer highs
summstart = pd.to_datetime('2020-06-22')
summend = pd.to_datetime('2020-09-01')
for curr_df, ax in [(la_df,axs[0]),(us_df,axs[1])]:
    dts = pd.to_datetime(curr_df['date'])
    summermask = (dts > summstart) & ( dts < summend)
    summerhigh = curr_df['NewDeathsRollAvg'][summermask].max()
    ax.axhline(summerhigh, label="Aprox. Summer high (%.0d)" % summerhigh, color='orange', alpha=0.4)
 
for ax in fig.axes:
    ax.legend()

fig.suptitle("Covid-19 La vs US New Deaths Daily")
fig.tight_layout()
plt.savefig("fig3.jpg")


# In[20]:


#Figure 4

#problem with the legend need to work on this
def plotme(df,ax,xticks, column, label, color = 'r', linestyle='-'):
    lines = []
    
    lines.append(ax.plot(df.index,df[column],color=color, label=label, linestyle=linestyle))
    ax.set_xticks(xticks)
    ax.set_xticklabels(df.iloc[xticks,0])
    
    
    drawindexofCDCdataloss(df,ax)
    lines.append(holidays(df,ax)) 
    
    return lines
    
    
fig, axs = plt.subplots(2, figsize=(15,17))

ax1lines = []
ax2lines = []

ax1lines.append( plotme(orleans_df,axs[0],orleans_ticks,'NewCasesPerCapita_ravg',"Orleans New Cases", color = 'r')[0][0])
ax1lines.append( plotme(ebr_df,axs[0],ebr_ticks,'NewCasesPerCapita_ravg',"Baton Rouge New Cases", color = 'orange')[0][0])
ax1lines.append( plotme(tamm_df,axs[0],orleans_ticks,'NewCasesPerCapita_ravg',"St. Tammany New Cases", color = 'purple')[0][0])
ax1lines.append( plotme(la2_df,axs[0],la2_ticks, 'NewCasesPerCapita_ravg',"La New Cases (LaDH)", color='b')[0][0])

ax1lines.append( plotme(la_df,axs[0],la_ticks, 'NewCasesPerCapita_ravg',"La New Cases (nytimes)", color='b', linestyle='--')[0][0])

ax1lines.append( plotme(us_df,axs[0],us_ticks, 'NewCasesPerCapita_ravg','US New Cases', color='g')[0][0])
ax1lines.append( plotme(world_df,axs[0],world_ticks,'NewCasesPerCapita_ravg',"World New Cases", color='m')[0][0])

ax1lines.append(ladhdatamax(axs[0]))

ax2lines.append( plotme(la_df,axs[1],la_ticks, 'NewDeathsPerCapita_ravg',"La New Deaths (nytimes)", color='b')[0][0])
ax2lines.append( plotme(us_df,axs[1],us_ticks, 'NewDeathsPerCapita_ravg','US New Deaths', color='g')[0][0])
ax2lines.append( plotme(world_df,axs[1],world_ticks,'NewDeathsPerCapita_ravg',"World New Daths", color='m')[0][0])

#ax2lines.append(ladhdatamax(axs[1]))  #no ladh data


#for col,ax in zip([('NewCasesPerCapita_ravg','New Cases'),('NewDeathsPerCapita_ravg', 'New Deaths')],((axs[0],ax1lines),(axs[1],ax2lines))):
    
#    ax[1].append( plotme(la_df,ax[0],la_ticks, col[0],"La %s" % col[1], color='b')[0][0])
#    ax[1].append( plotme(us_df,ax[0],us_ticks, col[0],'US %s' % col[1], color='g')[0][0])
#    ax[1].append( plotme(world_df,ax[0],world_ticks,col[0],"World %s" % col[1], color='m')[0][0])
    
axs[0].legend(handles=ax1lines,ncol=2)
axs[1].legend(handles=ax2lines)
    
    

axs[0].title.set_text('Orleans, EBR, Louisiana vs US vs World New Cases per 1million 7 day rolling avg')
axs[1].title.set_text('Louisiana vs US vs World New Deaths per 1million 7 day rolling avg')

axs[0].set_ylabel("New Cases per 1 million people")
axs[1].set_ylabel("New Deaths per 1 million people")

fig.suptitle("Covid-19 La vs US New Cases/Deaths per capita 7 day rolling avg ")
fig.tight_layout()
plt.savefig("fig4.jpg")


# In[21]:


#Figure 5

#problem with the legend need to work on this
def plotme(df,ax,xticks, column, label, color = 'r'):
    lines = []
    
    lines.append(ax.plot(df.index,df[column],color=color, label=label))
    ax.set_xticks(xticks)
    ax.set_xticklabels(df.iloc[xticks,0])
    
    drawindexofCDCdataloss(df,ax)
    lines.append(holidays(df,ax))
    #lines.append(LACFline(df,ax))   #I'm not sure the data for deaths and cases was effected by the adjustment
    
    yticks,yticklabels = getyTicks(df,column)
    ax.set_yticks(yticks)        
    ax.set_yticklabels(yticklabels)
    ax.set_ylabel(label,color=color)
    ax.tick_params(axis='y',color=color,labelcolor=color)
    
    return lines
    
    
fig, axs = plt.subplots(4, figsize=(15,15))

idx = 0
for col, label in zip(['NewCasesRollAvg','NewDeathsRollAvg'], ['New Cases','New Deaths']):
    for compare_place, label_place in zip([(us_df,us_ticks),(world_df,world_ticks)], ['US','World']):
        line1 = plotme(la_df,axs[idx],la_ticks, col,"La %s" % label, color='b')
        axsY2 = axs[idx].twinx()
        line2 = plotme(compare_place[0],axsY2,compare_place[1], col,"%s %s" % (label_place, label), color='g')
        axs[idx].legend(handles=[line1[0][0],line2[0][0]],loc='upper left')
        axs[idx].title.set_text("Louisiana (nytimes data) vs %s %s 7 day rolling avg"%(label_place,label))
        idx += 1
    
fig.suptitle("Covid-19 Louisiana vs * New Cases 7 day rolling avg Compared (separatly scaled)")
fig.tight_layout()
plt.savefig("fig5.jpg")


# In[22]:


#Figure 5

#problem with the legend need to work on this
def plotme(df,ax,xticks, column, label, color = 'r'):
    lines = []
    
    lines.append(ax.plot(df.index,df[column],color=color, label=label))
    ax.set_xticks(xticks)
    ax.set_xticklabels(df.iloc[xticks,0])
    
    drawindexofCDCdataloss(df,ax)
    lines.append(holidays(df,ax))
    #lines.append(LACFline(df,ax))   #I'm not sure the data for deaths and cases was effected by the adjustment
    lines.append(ladhdatamax(ax))
    
    yticks,yticklabels = getyTicks(df,column)
    ax.set_yticks(yticks)        
    ax.set_yticklabels(yticklabels)
    ax.set_ylabel(label,color=color)
    ax.tick_params(axis='y',color=color,labelcolor=color)
    ax.set_xlim(xmax=xdatalen)
    
    return lines
    
    
fig, axs = plt.subplots(5, figsize=(15,17))

idx = 0
col = 'NewCasesRollAvg'
label = 'New Cases'
for compare_place, label_place in zip([(ebr_df,ebr_ticks,'orange'),(tamm_df,tamm_ticks,'purple'), (la2_df,la2_ticks,'b'), (us_df,us_ticks,'g'),(world_df,world_ticks,'m')], ['Baton Rouge', 'St. Tammany','Louisiana (LaDH datasource)','US','World']):
    line1 = plotme(orleans_df,axs[idx],orleans_ticks, col,"Orleans %s" % label, color='r')
    axsY2 = axs[idx].twinx()
    line2 = plotme(compare_place[0],axsY2,compare_place[1], col,"%s %s" % (label_place, label), color=compare_place[2])
    axs[idx].legend(handles=[line1[0][0],line2[0][0]])
    axs[idx].title.set_text("New Orleans vs %s %s 7 day rolling avg"%(label_place,label))
    idx += 1

fig.suptitle("Covid-19 La vs US vs World New Cases/Deaths 7 day rolling avg Compared (separatly scaled, (New Orleans/EBR uses collection date)")
fig.tight_layout()
plt.savefig("fig8.jpg")


# In[23]:


weeks = []
day = mindate
deltaweek = timedelta(days=7)
enddate = datetime.now() - deltaweek
while day <= enddate:
    weeks.append(day)
    day = day + deltaweek

      
def getWeeklyDF(df):
    
    df.index = df['date'].map( lambda x: pd.Timestamp(x))
    dfmaxdate = pd.to_datetime(df.index.max())
    enddate = dfmaxdate - timedelta(days=7)
    
    newdf = pd.DataFrame(index = weeks)
    
    columns = ['NewCases', 'NewDeaths']
    for column in columns:
        if column in df.columns:
        
            totals = []
            deltaweek = timedelta(days=7)
            for week in weeks:
                if week <= enddate:
                    mask = (df.index < week + deltaweek) & (df.index >= week)
                    totals.append(np.nansum(df[column][mask]))
                else:
                    totals.append(np.nan)
                
            newdf[column] = pd.Series(totals,index=weeks)
            newcol = column + 'RollAvg'
            newdf[newcol] = rollingavg(newdf[column],3)
            newdf[newcol] = np.where(newdf[column] > 0, newdf[newcol], np.nan)

    return newdf

orleans_df_weekly = getWeeklyDF(orleans_df) #pd.DataFrame(orleans_df.groupby('week')['NewCases'].sum())
ebr_df_weekly = getWeeklyDF(ebr_df) #pd.DataFrame(ebr_df.groupby('week')['NewCases'].sum())
tamm_df_weekly = getWeeklyDF(tamm_df)#pd.DataFrame(tamm_df.groupby('week')['NewCases'].sum())
la2_df_weekly = getWeeklyDF(la2_df)#pd.DataFrame(la2_df.groupby('week')['NewCases'].sum())
la_df_weekly = getWeeklyDF(la_df)#la_df.groupby('week')[['NewCases','NewDeaths']].sum()
us_df_weekly = getWeeklyDF(us_df)#.groupby('week')[['NewCases','NewDeaths']].sum()
world_df_weekly = getWeeklyDF(world_df)#.groupby('week')[['NewCases','NewDeaths']].sum()

print(weeks[-1])
world_df_weekly.tail()
#us_df_weekly.tail(2)


# In[24]:


#Figure 6

def plotme(df,ax, title = "Graphs", color = 'y', linecolor = 'r'):
    ax.bar(df.index,df['NewCases'], color=color, label='weekly new cases', width=6.5)
    ax.plot(df.index,df['NewCasesRollAvg'], color=linecolor, label='3 week rolling avg')
    
    ax.set_title(title)
    ax.set_xticks(df.index)
    ax.xaxis.set_major_formatter(mdates.DateFormatter('%b %d'))
    ax.tick_params(axis='x',rotation=90)
    ax.set_xlim((weeks[0]-timedelta(days=4),weeks[-1] + timedelta(days=5)))
    ax.legend()
      
fig, axs = plt.subplots(6, figsize=(15,18))
fig.suptitle("Weekly New Cases")

plotme(orleans_df_weekly,axs[0], title="Orleans")
plotme(ebr_df_weekly,axs[1],title="Baton Rough")
plotme(tamm_df_weekly,axs[2],title="St. Tammany")


axs[3].plot(la2_df_weekly.index, la2_df_weekly['NewCasesRollAvg'], linestyle='--', color='r', label='3 week rolling avg (LaDH)')
plotme(la_df_weekly,axs[3], title="Louisiana (nytimes data)")
plotme(us_df_weekly,axs[4], title="US")
plotme(world_df_weekly,axs[5],title="World")

fig.tight_layout()

plt.savefig("fig6.jpg")


# In[25]:


#Figure 7

def plotme(df,ax, title = "Graphs", color = 'y', linecolor='r'):
    ax.bar(df.index,df['NewDeaths'], color=color, label='weekly new deaths', width=6.5)
    ax.plot(df.index,df['NewDeathsRollAvg'], color=linecolor, label='3 week rolling avg')
    
    ax.set_title(title)
    ax.set_xticks(df.index)
    ax.xaxis.set_major_formatter(mdates.DateFormatter('%b %d'))
    ax.tick_params(axis='x',rotation=90)
    ax.set_xlim((weeks[0]-timedelta(days=4),weeks[-1] + timedelta(days=5)))
    ax.legend()
    
fig, axs = plt.subplots(3, figsize=(15,12))
fig.suptitle("Weekly New Deaths")

plotme(la_df_weekly,axs[0], title="Louisiana (nytimes data)")
plotme(us_df_weekly,axs[1], title="US")
plotme(world_df_weekly,axs[2],title="World")

fig.tight_layout()

plt.savefig("fig7.jpg")


# In[ ]: