How to specify x and y in the dataframe groupby plot

Question

I have a dataframe which contains essentially 3 columns(I have many columns but the main idea is presented here):
df["Label","Data1","Data2"]

df = pd.DataFrame({'Label': [1, 1, 1, 2, 2, 3, 3, 3, 4, 4],
                   'Data1': [0.1, 0.01, 0.15, 0.3, 0.35, 0.38, 0.44, 0.45, 0.8, 0.88],
                   'Data2': [1.4, 1.4, 1.6, 2.3, 2.5, 3.6, 3.8, 3.9, 7.3, 7.7]})

I am wondering if there is a way to plot Data1 vs Data2, groupby(Label) and give colors based on the groupby?
I tried something like:

df.groupby("Label")[["Data1","Data2"]].plot(marker='.',subplots=False,ax=plt.gca())

But the figure is NOT about Data1 vs Data2.

Also I would like to know if this could be done in Matplotlib.

Asked By: yangyang

||

Source

Answer 1

Are you trying to just plot a scatter plot with Data1 on x-axis and Data2 on y-axis, with the colours of each dot being related to the Label column? If so, you don’t really need to groupby and instead can just do:

Setup

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt

df = pd.DataFrame({'Label': [1, 1, 1, 2, 2, 3, 3, 3, 4, 4],
                   'Data1': [0.1, 0.01, 0.15, 0.3, 0.35, 0.38, 0.44, 0.45, 0.8, 0.88],
                   'Data2': [1.4, 1.4, 1.6, 2.3, 2.5, 3.6, 3.8, 3.9, 7.3, 7.7]})

For string label values

I struggled a bit with the string values, there might be a cleaner way but this is the way that worked for me:

Method 1 default colours

fig,ax = plt.subplots()

#default colour palette
prop_cycle = plt.rcParams['axes.prop_cycle'] 
#list of colours
colors = prop_cycle.by_key()['color'] 

i=0 #iterator

#iterate over each label value
for label in df['Label'].unique():

    #get datapoints for that label
    x = df[df['Label']==label]['Data1']
    y = df[df['Label']==label]['Data2']
    
    #specify color and label (for legend)
    ax.scatter(x,y,color=colors[i],label=label)
    i+=1
    
#show legend
ax.legend()
plt.show()

Method 2 color palette

from matplotlib import cm

fig,ax = plt.subplots()

n = df['Label'].nunique() #num of categs

#colours for rainbow palette (can use other palette)
color = iter(cm.rainbow(np.linspace(0, 1, n))) 

for label in df['Label'].unique():
    c = next(color) #change colour
    x = df[df['Label']==label]['Data1']
    y = df[df['Label']==label]['Data2']
    
    ax.scatter(x,y,color=c,label=label)
    
ax.legend()
plt.show()

For numerical label values:

plt.scatter(df['Data1'],df['Data2'],c=df['Label'])

The c = is essentially all you need.

You can also specifycmap='plasma' for example, to change the colour palette. List of colour maps can be found on matplotlib docs

Answered By: mosefaq