Pandas plot bar order categories

问题

I have a dataset with a categorical variable that contains three unique values, "low", "medium" and "high":

df.CatVar.value_counts()
Out[93]: 
Medium    35832
Low       25311
High      12527
Name: CatVar, dtype: int64

I am trying to plot the number of unique values as a bar-plot. However, the following code gives me the bars in the order ["Medium", "Low", "High"]

df.CatVar.value_counts().plot(kind="bar")

How do I change the order of the bars in the plot?

回答1:

There are 2 possible solutions - change order of index before plot - by reindex or loc:

df.CatVar.value_counts().reindex(["Low", "Medium", "High"]).plot(kind="bar")

df.CatVar.value_counts().loc[["Low", "Medium", "High"]].plot(kind="bar")

Or use ordered categorical, so after value_counts get order by categories parameter:

df.CatVar = pd.Categorical(df.CatVar, categories=["Low", "Medium", "High"], ordered=True)
df.CatVar.value_counts(sort=False).plot(kind="bar")

---

Sample:

df = pd.DataFrame({'CatVar':['Low','Medium','Low','Low','Medium','High']})
print (df)
   CatVar
0     Low
1  Medium
2     Low
3     Low
4  Medium
5    High

df.CatVar.value_counts().reindex(["Low", "Medium", "High"]).plot(kind="bar")

回答2:

The following code solved my problem:

df.CatVar.value_counts()[['Low', 'Medium', 'High']].plot(kind="bar")

回答3:

If you do not mind using seaborn, you can use countplot and it has parameter to pass the order:

import seaborn as sns

df = pd.DataFrame({'CatVar':['Low','High','Low','Low','Medium']})
sns.countplot(x='CatVar', data=df, order=['Low', 'Medium', 'High']);

来源：https://stackoverflow.com/questions/50553698/pandas-plot-bar-order-categories