按类别顺序对观察数进行排序

Question

我有以下代码，该代码生成一个Seaborn剥离图，然后将观测值的数量写在每个类别下。如果我在stripplot（）调用中指定了不同的类别顺序，则数字会乱序。

我需要帮助找出一种方法来对我的nob系列进行排序，以使数字与cat_order中类别的顺序匹配。有没有一种优雅的方法可以做到这一点？

# import libraries
import pandas as pd
import numpy as np
import seaborn as sns
import matplotlib.pyplot as plt

# create some fake data

df = pd.DataFrame(data=None,columns=['Fuel','MW'])
df = df.astype(dtype={'Fuel':'category', 'MW':'float'})
rnd = np.random.RandomState(1)
mymean = 4 # starting mean value for use in the distribution below
mysigma = 0.8 # sigma value for use in the distribution below
myfuels = ['Bio','Coal','Hydro','Gas','Oil','Other','Solar','Wind'] # categories
mysize = 10 
for myfuel in myfuels:
    myMWs = list(rnd.lognormal(mean=mymean,sigma=mysigma,size=mysize)) # random generated data
    newdata = pd.DataFrame({'Fuel': [myfuel]*mysize,
                           'MW':  myMWs})
    df = df.append(newdata,
                   ignore_index=True
                  )    
    mymean = mymean + 0.4 # increment to have different distributions
    mysize = mysize + 3

# see distribution of created data

# Set up figure and axes
fig1, ax1 = plt.subplots(nrows=1, ncols=1, figsize=(14,5))

sns.set_style('darkgrid')

# define category order (same as myfuels, except Other is at end)
cat_order = ['Bio','Coal','Hydro','Gas','Oil','Solar','Wind','Other']

sns.stripplot(x = 'Fuel',
               y = 'MW',
               data=df,
               order=cat_order,
               jitter=0.35,
               size=2.5,
               ax = ax1,
               palette='dark'
              )   

ax1.set_ylim([-400,5500])

# add tags for number of observations in each category
nobs = df.groupby(['Fuel'])['MW'].agg(['count'])['count'] #Is there a more elegant form of this line?
myv = -280 # get this value through inspection and trial and error
myh = 0 # first category is always at horiz position = 0
for item in nobs:
    if myh == 0:
        plt.text(myh, myv, 'n = ' + str(item), ha='center', fontsize=9)
    else:
        plt.text(myh, myv, item, ha='center', fontsize=9)
    myh = myh + 1 # move one unit to the right for each category

plt.show()

这是当前输出（带有错误的观察次数指示）。

Answer 1

您要按照用于绘制带状图的类别顺序为nobs系列重新编制索引。

nobs = df.groupby(['Fuel'])['MW'].agg(['count'])['count'].reindex(cat_order)