I'm comparing several histogram together, to evaluate the variance over the years:
The problem is that, while it shows the variance, it is hard to discern one type from another. I'm wondering what would be a better way to make this plot?
Add a comment
|
1 Answer
Those bars are not really stacked but overlapped. You can stack them by doing:
import matplotlib.pyplot as plt
import numpy as np
mu, sigma = 0,100
x1 = mu + sigma*np.random.randn(1000)
x2 = mu + sigma*np.random.randn(1000)
x3 = mu + sigma*np.random.randn(1000)
#Stack the data
plt.figure()
plt.hist([x1,x2,x3], bins=30, stacked=True, normed = True)
plt.show()
, which will result in this:
However to perceive a quantile variability in different distributions the boxplot is usually more adequate:
plt.boxplot([x1,x2,x3])
plt.show()
, results in:
You can also take a look into plots such as Probability Plot, or StackPlot. If you really need to compare the variance value (and not just the distribution globally) you can plot the lines for variance (or standard deviation if the order of magnitude is big) over your plots.
4 Comments
ZK Zhao
My mistake. I wrongly used
stack. It should be overlap. The boxplot would lose the histogram shape, which I think is really important here.
armatita
Than use the Violin Plot (matplotlib.org/examples/statistics/violinplot_demo.html). What features are you trying to compare (variance is just a number)?
ZK Zhao
No, no,no. I don't need to compare this specfic here, e.g. variance etc. . I just want to compare the histogram. The problem is that overlapping made them hard to look at. The half-transparent setting (
alpha) makes it hard to discern one from another.armatita
You have an argument that allows you to remove the fill from the bars in matplotlib histogram. Try adding histtype='step' as argument. It should only show you the contours of the histogram (micropore.wordpress.com/2011/11/15/…). Also take a look at this: matplotlib.org/examples/statistics/…