Overlay a line function on a scatter plot
Question:
My challenge is to overlay a custom line function graph over a scatter plot I already have, the code looks like follows:
base_beta = results.params
X_plot = np.linspace(0,1,400)
g = sns.FacetGrid(data, size = 6)
g = g.map(plt.scatter, "usable_area", "price", edgecolor="w")
Where base_beta is only a constant, and then one coefficient. Basically, I want to overlay a function that plots a line y = constant + coefficient * x
I tried to overlay a line using this but it did not work.
g = g.map_dataframe(plt.plot, X_plot, X_plot*base_beta[1]+base_beta[0], 'r-')
plt.show()
The current scatter plot looks like so:
Can any one help me with this?
–ATTEMPT 1
base_beta = results.params
X_plot = np.linspace(0,1,400)
Y_plot = base_beta [0] + base_beta[1]*X_plot
g = sns.FacetGrid(data, size = 6)
g = g.map(plt.scatter, "usable_area", "price", edgecolor="w")
plt.plot(X_plot, Y_plot, color='r')
plt.show()
Resulted in the same graph but no line:
Answers:
You can just call plt.plot to plot a line over the data.
import numpy as np
import matplotlib.pyplot as plt
import pandas as pd
import seaborn as sns
data = pd.DataFrame()
data['usable_area'] = 5*np.random.random(200)
data['price'] = 10*data['usable_area']+10*np.random.random(200)
X_plot = np.linspace(0, 7, 100)
Y_plot = 10*X_plot+5
g = sns.FacetGrid(data, height = 6)
g = g.map(plt.scatter, "usable_area", "price", edgecolor="w")
plt.plot(X_plot, Y_plot, color='r')
plt.show()
Produces:
You can also overlay a Seaborn plot over the data, given you have the points that make up that line (below, I call them x_pred and y_pred):
fig, ax = plt.subplots(figsize=(11, 8.5))
sns.scatterplot(x='M2NS_PC1', y='FII5', data=ir_ms, ax=ax)
ax.axhline(y=0, color='k', linewidth=1) # added because i want the origin
ax.axvline(x=0, color='k', linewidth=1)
fitted = sm.ols(formula='FII5 ~ M2NS_PC1', data=ir_ms).fit(cov_type='HC3')
x = ir_ms['M2NS_PC1']
x_pred = np.linspace(x.min() - 1, x.max() + 1, 50)
y_pred = fitted.predict(exog=dict(M2NS_PC1=x_pred))
sns.lineplot(x=x_pred, y=y_pred, ax=ax)
Then, just plot it all on the same axis.
- It is now recommended to use figure-level functions like
seaborn.relplot or seaborn.regplot instead of directly using seaborn.FacetGrid.
- Tested in
python 3.8.12, pandas 1.3.3, matplotlib 3.4.3, seaborn 0.11.2
Sample Data and Imports
import numpy as np
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
# create a dataframe with sample x and y
np.random.seed(365)
x = 5*np.random.random(200)
df = pd.DataFrame({'x': x, 'y': 10*x+10*np.random.random(200)})
# add custom line to the dataframe
base_beta = [10, 5]
df['y_line'] = base_beta[0] + base_beta[1]*df.x
display(df.head())
x y y_line
0 4.707279 50.634968 33.536394
1 3.208014 33.890507 26.040068
2 3.423052 37.853276 27.115262
3 2.942810 29.899257 24.714052
4 2.719436 36.932170 23.597180
Add Custom Line to Scatter Plot
sns.relplot with .map or .map_dataframe
- Apply an axes-level plotting function (e.g.
sns.lineplot) to each facet of the figure-level plot.
- seaborn: Building structured multi-plot grids
ax = sns.relplot(kind='scatter', x='x', y='y', data=df, height=3.5, aspect=1.5)
ax.map_dataframe(sns.lineplot, 'x', 'y_line', color='g')
sns.scatterplot with sns.lineplot
- Plot two axes-level plots to the same figure.
fig, ax = plt.subplots(figsize=(6, 4))
p1 = sns.scatterplot(data=df, x='x', y='y', ax=ax)
p2 = sns.lineplot(data=df, x='x', y='y_line', color='g', ax=ax)
Regression Line to a Scatter Plot
- For a regression line
- Use
seaborn.lmplot for figure-level regression plot
- Use
seaborn.regplot for an axes-level regression plot.
sns.lmplot
g = sns.lmplot(data=df, x='x', y='y', line_kws={'color': 'g'}, height=3.5, aspect=1.5)
sns.regplot
ax = sns.regplot(data=df, x='x', y='y', line_kws={'color': 'g'})
My challenge is to overlay a custom line function graph over a scatter plot I already have, the code looks like follows:
base_beta = results.params
X_plot = np.linspace(0,1,400)
g = sns.FacetGrid(data, size = 6)
g = g.map(plt.scatter, "usable_area", "price", edgecolor="w")
Where base_beta is only a constant, and then one coefficient. Basically, I want to overlay a function that plots a line y = constant + coefficient * x
I tried to overlay a line using this but it did not work.
g = g.map_dataframe(plt.plot, X_plot, X_plot*base_beta[1]+base_beta[0], 'r-')
plt.show()
The current scatter plot looks like so:
Can any one help me with this?
–ATTEMPT 1
base_beta = results.params
X_plot = np.linspace(0,1,400)
Y_plot = base_beta [0] + base_beta[1]*X_plot
g = sns.FacetGrid(data, size = 6)
g = g.map(plt.scatter, "usable_area", "price", edgecolor="w")
plt.plot(X_plot, Y_plot, color='r')
plt.show()
Resulted in the same graph but no line:
You can just call plt.plot to plot a line over the data.
import numpy as np
import matplotlib.pyplot as plt
import pandas as pd
import seaborn as sns
data = pd.DataFrame()
data['usable_area'] = 5*np.random.random(200)
data['price'] = 10*data['usable_area']+10*np.random.random(200)
X_plot = np.linspace(0, 7, 100)
Y_plot = 10*X_plot+5
g = sns.FacetGrid(data, height = 6)
g = g.map(plt.scatter, "usable_area", "price", edgecolor="w")
plt.plot(X_plot, Y_plot, color='r')
plt.show()
Produces:
You can also overlay a Seaborn plot over the data, given you have the points that make up that line (below, I call them x_pred and y_pred):
fig, ax = plt.subplots(figsize=(11, 8.5))
sns.scatterplot(x='M2NS_PC1', y='FII5', data=ir_ms, ax=ax)
ax.axhline(y=0, color='k', linewidth=1) # added because i want the origin
ax.axvline(x=0, color='k', linewidth=1)
fitted = sm.ols(formula='FII5 ~ M2NS_PC1', data=ir_ms).fit(cov_type='HC3')
x = ir_ms['M2NS_PC1']
x_pred = np.linspace(x.min() - 1, x.max() + 1, 50)
y_pred = fitted.predict(exog=dict(M2NS_PC1=x_pred))
sns.lineplot(x=x_pred, y=y_pred, ax=ax)
Then, just plot it all on the same axis.
- It is now recommended to use figure-level functions like
seaborn.relplotorseaborn.regplotinstead of directly usingseaborn.FacetGrid. - Tested in
python 3.8.12,pandas 1.3.3,matplotlib 3.4.3,seaborn 0.11.2
Sample Data and Imports
import numpy as np
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
# create a dataframe with sample x and y
np.random.seed(365)
x = 5*np.random.random(200)
df = pd.DataFrame({'x': x, 'y': 10*x+10*np.random.random(200)})
# add custom line to the dataframe
base_beta = [10, 5]
df['y_line'] = base_beta[0] + base_beta[1]*df.x
display(df.head())
x y y_line
0 4.707279 50.634968 33.536394
1 3.208014 33.890507 26.040068
2 3.423052 37.853276 27.115262
3 2.942810 29.899257 24.714052
4 2.719436 36.932170 23.597180
Add Custom Line to Scatter Plot
sns.relplot with .map or .map_dataframe
- Apply an axes-level plotting function (e.g.
sns.lineplot) to each facet of the figure-level plot. - seaborn: Building structured multi-plot grids
ax = sns.relplot(kind='scatter', x='x', y='y', data=df, height=3.5, aspect=1.5)
ax.map_dataframe(sns.lineplot, 'x', 'y_line', color='g')
sns.scatterplot with sns.lineplot
- Plot two axes-level plots to the same figure.
fig, ax = plt.subplots(figsize=(6, 4))
p1 = sns.scatterplot(data=df, x='x', y='y', ax=ax)
p2 = sns.lineplot(data=df, x='x', y='y_line', color='g', ax=ax)
Regression Line to a Scatter Plot
- For a regression line
- Use
seaborn.lmplotfor figure-level regression plot - Use
seaborn.regplotfor an axes-level regression plot.
- Use
sns.lmplot
g = sns.lmplot(data=df, x='x', y='y', line_kws={'color': 'g'}, height=3.5, aspect=1.5)
sns.regplot
ax = sns.regplot(data=df, x='x', y='y', line_kws={'color': 'g'})