Overview and tips of seaborn with statistical data visualization
seaborn: Python's data visualization library. High level API. Beautiful.
0. How to use with matplotlib
-
Use
seabornif it is prepared byseaborn. If not, draw withmatplotlib. -
Even if you draw with
matplotlib, it is better to useseabornfor the style (it is set just by importing)
1. Style settings
I often use default or "white" for style. Adjust font_scale by looking at the figure.
import seaborn as sns
sns.set(style="white", font_scale=1.3, palette="muted", color_codes=True)
2. Check if you can do what you want with seaborn
View the gallery page of the official doc. (Since the functions that can be used are limited to the following, if you can imagine what you can do with the function name, you do not have to look at the link. You can check the details of the arguments with the help function of ipython.)
Visualization of distribution
function
- distplot : 1-dim.
- jointplot : 2-dim.
- pairplot : k-dim.
- heatmap: Correlation matrix
argument --kde: Kernel density estimation --hist: Histogram --Hue: Distribution estimation by condition
Regression
function
- lmplot: Regression with confidence intervals
argument --order: Polynomial regression --logistic: Logistic regression --hue: Conditional regression (col, row)
Categorical data
function
- stripplot, swarmplot: Scatter plot
- boxplot, violinplot: Distribution visualization
- barplot, countplot, pointplot: Visualization of statistical properties
- factorplot (hue ='key'): Conditional visualization of statistical properties
3. Other
Read the dataset provided by seaborn
#dataset list: https://github.com/mwaskom/seaborn-data
sns.load_dataset('titanic')
Shaft removal
sns.despine() #Top and right
sns.despine(left=True) #Left too
color palettes
#Current color map
cmap_current = sns.color_palette()
sns.palplot(cmap_current)
#New color map
#reference: http://seaborn.pydata.org/tutorial/color_palettes.html
#Example: key = "set1" (categorical), "Blues" (sequential)
num = 8
cmap = sns.color_palette(key, num)
sns.palplot(cmap)
Recommended Posts