gini-calculation.py

# Gini function as PyGini package
def gini(arr, eps=1e-8):
    '''
    Reference: PyGini (I owe you a beer @o_guest)
    https://github.com/mckib2/pygini/blob/master/pygini/gini.py
    
    Calculate the Gini coefficient of a numpy array.
    Notes
    -----
    Based on bottom eq on [2]_.
    References
    ----------
    .. [2]_ http://www.statsdirect.com/help/
            default.htm#nonparametric_methods/gini.htm
    '''

    # All values are treated equally, arrays must be 1d and > 0:
    arr = np.abs(arr).flatten() + eps

    # Values must be sorted:
    arr = np.sort(arr)

    # Index per array element:
    index = np.arange(1, arr.shape[0]+1)

    # Number of array elements:
    N = arr.shape[0]

    # Gini coefficient:
    return(np.sum((2*index - N - 1)*arr))/(N*np.sum(arr))


def get_gini_df(df):
    """Generate DF with Gini Index

    Parameters
    ----------
    df : Pandas Dataframe
      Dataframe with Brasileirão data

    Returns
    -------
    gini_df : Pandas Dataframe
      Returns a Pandas Dataframe with the year, team and gini index

    """    
    gini_per_year = []
    
    for year in df['year'].unique():
        championship_index = gini(np.array(df[df['year'] == year]['points']))
        champion = (df[(df['year'] == year) & (df['position'] == 1)]['team'])
        gini_per_year.append((year, champion.values[0], round(championship_index, 4)))
        
    gini_df = pd.DataFrame(gini_per_year)
    gini_df.columns = ['year', 'team', 'gini']
    
    # Indexing the date field for graph it smoothly
    gini_df.set_index('year', inplace=True)

    return gini_df
  
  
gini_df = get_gini_df(df_brasileirao)
gini_df.sort_values(by=['gini'], ascending=True)