Fork me on GitHub!

### Reasoning¶

My maternal grandparents are both quite old and have their share of health issues requiring constant monitoring. We annotate daily both their blood pressure and pulse level and my grandma's blood-sugar level since she has diabetes.

All this data is unfortunately kept analogically by mere pen and paper, in order to ingest the data I have to manually copy the values into a Google spreadsheet (as to make it readily available on the cloud) and then export is as a .csv file to analyze.

I have chosen Python, over R, as I want to get better at it, for the data visualization aspect I have chosen Altair since I like its declarative style and its novelty make me want to try and master it

As of now I am interested in exploring the data, once I collect more precise information on their medication I would like to try my hand at some machine learning training

2017-05-28: As of today I only have my grandpa data for the 2017

### TODO¶

• add more data on pressure [insert values for 2016]
• see if there's a way to normalize the dimensions of the histogram
• add more data on blood-pressure drugs taken daily as to check for a correlation
• Change the color of the type
• Rearrange the level the position of the histograms
• Add an horizontal line representing the threshold-risk value
In [14]:
import pandas as pd
import altair

In [15]:
DATA_URL = 'https://docs.google.com/spreadsheets/d/1r2YybTa9dCHZWz9NvTI3zN4pfthryyJAvjG2BPoOwr8/pub?gid=0&single=true&output=csv'  # Now it will be easy to access
df_raw.drop(['medication', 'lasix'], axis=1, inplace=True)  # Until I have more information on its medication and lasix intake I am not factoring in these
df_raw.dropna(inplace=True)  # Some value are missing, we will drop them


## Transforming the dataframe¶

Note how our dataframe would not allow for multiple simultaneous representations. However we can Pivot it to obtain a df we can work with.

In [16]:
df = pd.melt(df_raw, id_vars=['date'], value_vars=['systolic', 'diastolic', 'pulse'], var_name='type', value_name='measurement')

In [31]:
chart = altair.Chart(df).mark_line().encode(
x = 'date', y = 'measurement', color='type'
)
chart.configure_cell(height=350)

In [18]:
systolic = df[df['type'] == 'systolic'].drop(['type'], axis=1)
systolic.name = 'Systolic Pressure'
print(systolic.mean())
print(systolic.std())

measurement    161.367647
dtype: float64
measurement    20.298443
dtype: float64

In [19]:
diastolic = df[df['type'] == 'diastolic'].drop(['type'], axis=1)
diastolic.name = 'Diastolic Pressure'
print(diastolic.mean())
print(diastolic.std())

measurement    81.536765
dtype: float64
measurement    8.949991
dtype: float64

In [20]:
pulse = df[df['type'] == 'pulse'].drop(['type'], axis=1)
pulse.name = 'Pulse'
print(pulse.mean())
print(pulse.std())

measurement    65.139706
dtype: float64
measurement    5.045237
dtype: float64


## Testing the distribution¶

In [21]:
from scipy.stats import normaltest


We now construct a function to test whether the pressure is normally distributed or not.

Note on scipy.stats.normaltest:

Tests whether a sample differs from a normal distribution.

This function tests the null hypothesis that a sample comes from a normal distribution. It is based on Dâ€™Agostino and Pearsonâ€™s [R614], [R615] test that combines skew and kurtosis to produce an omnibus test of normality.

Source

In [22]:
def is_normally_distributed(array):
'''
We pass the "polished" dataframe for both type  of pressure (and pulse),
we then apply the normaltest function on the pressure column
'''
_, p = normaltest(array['measurement'])  # I am interested only in the p-value
print(p)
if p < 0.055:  # P value treshold to test the null hypothesis
print('Normally Distributed\n')
else:
print('Not normally distributed\n')

In [23]:
measurements_type = [systolic, diastolic, pulse]

for i in measurements_type:
print(i.name)
is_normally_distributed(i)

Systolic Pressure
0.000837359797293
Normally Distributed

Diastolic Pressure
0.00303140068803
Normally Distributed

Pulse
2.76374279722e-25
Normally Distributed



## Visualizing the distribution¶

In [24]:
c = altair.Chart(df[df['type']!='pulse']).mark_bar().encode(
altair.X('measurement', bin=altair.Bin(step=5), title='Measurement'),  # I choose step=5 to have a better representation of the distribution especially fot Max and Medium Pressure
altair.Y('*', aggregate='count', type='quantitative'),
color='type',
column=altair.Column('type',
title='Blood pressure')
)
c.configure_cell(height=350, width=350)