##identify pulses

#pad dataset by sampling interval in case there are any missing values
sample_interval = xmin

df['timestamp']

df.set_index('timestamp', inplace=True)

df_padded = df.resample('20min').asfreq()


#find 150 rolling mean
df['atnrm150']=df['atn'].rolling(150, min_periods=23).mean()

#Find raw, 24H rolling mean, 3-d rolling mean, 5-d rolling mean

df['rm24h'] = df['val'].rolling('24h').mean()

df['rm3d'] = df['val'].rolling('3d').mean()

df['rm5d'] = df['val'].rolling('5d').mean()

#find mean +2SD threshold for those

stats = {}
for col in columns_to_calc:
    mean = df[col].mean()
    sd = df[col].std()
    stats[f'{col}_mean_plus_2sd'] = mean + (2 * sd)

atnlimit_raw = stats['atn_mean_plus_2sd']
atnlimit_24H = stats['rm24h_mean_plus_2sd']
atnlimit_3d = stats['rm3d_mean_plus_2sd']
atnlimit_5d = stats['rm5d_mean_plus_2sd']

#Value exceeds dataset mean + 2SD for at least 3 of the following datasets: raw values, 24H rolling mean, 3-day rolling mean, and 5-day rolling mean. 


df['Pulsecheck_atnraw']=np.where(df['atn'] >= atn_mean_plus_2sd, 1, 0)
df['Pulsecheck_rm24h']=np.where(df['rm24h'] >= rm24h_mean_plus_2sd, 1, 0)
df['Pulsecheck_rm3d']=np.where(df['rm3d'] >= rm3d_mean_plus_2sd, 1, 0)
df['Pulsecheck_rm5d']=np.where(df['rm5d'] >= rm5d_mean_plus_2sd, 1, 0)

#“Pulse” starts with the first time this condition is met, and ends when the following 24 hours do not meet this condition. 
df['Pulsecheckcount'] = df[['Pulsecheck_atnraw', 'Pulsecheck_rm24h', 'Pulsecheck_rm3d', 'Pulsecheckrm5d']].sum(axis=1)
 
df['Pulsecheck_count']=np.where(df['Pulsecheckcount'] >= 3, 1, 0)

# Returns True if any value in the last 24h was > threshold
df['Pulsecheck_24prior'] = df['Pulsecheck_count'].rolling(window='24h').max() >= 1