熊猫-列的累计比例

Question

I have a dataframe column that is either 0, 1, or 2. I want to plot the relative proportions over time in a stacked bar chart. Eg if the values are:

0 1 2 2 0 0 1

Then the 0 - % value would be (rounded to 1.dp):

100 50 33 25 40 50 42

And the 1 - % value would be (again rounded to 1.dp):

0 50 33 25 20 33  

I would want the 0, 1 and 2 proportions to all be stacked in a single bar, showing how the relative proportion changes over time.

Answer 1

Okay first i need to make the obligatory complain that you did not provide any attempts you made so far, shame on you ;).

Nevertheless, let us help you. First one should break this task into little steps. We need to: 1. Create Indicator-Column for each value 2. We need the Cumsum for each of these 3. Divide it by the respective row number (+1 since indexing starts at 0 ) 4. Plot this beautiful thing

my attempt would be - not beautful , but brute force coding - :

# Create Example Data
df = pd.DataFrame(np.random.randint(0,4, 10), columns=['A'] )

# The function to make it one go 
def create_rolling_stack(df, column):

    # Create the Indicators also called OneHotEncoding or DummyEncoding
    dum = pd.get_dummies(df[column])

    # build cumsum 
    cums = dum.cumsum()

    # reset index
    cums = cums.reset_index(drop=True)

    # create the divisior 
    cums['div'] = cums.index.values +1

    # ugly but divde each column by the respective row number
    for col in cums.columns:
        cums[col] = cums[col]/cums['div'] 


    cums = cums.drop('div', axis = 1)
    # Plot this awesome thing, note that stacked is set to True 
    cums.plot(kind= 'bar', stacked = True )
    plt.show() 

Hope it helps