定量数据本质上是数值，应该是衡量某样东西的数量。
定性数据本质上是类别，应该是描述某样东西的性质。
全部的数据列如下，其中既有定性列也有定量列；
import pandas as pdpd.options.display.max_columns = None
pd.set_option('expand_frame_repr', False)
salary_ranges = pd.read_csv('./data/Salary_Ranges_by_Job_Classification.csv')
print(salary_ranges.head())
# SetID JobCode Eff Date SalEndDate SalarySetID SalPlan Grade Step BiweeklyHighRate BiweeklyLowRate UnionCode ExtendedStep PayType
# 0 COMMN 109 07/01/2009 12:00:00 AM 06/30/2010 12:00:00 AM COMMN SFM 0 1 $0.00 $0.00 330 0 C
# 1 COMMN 110 07/01/2009 12:00:00 AM 06/30/2010 12:00:00 AM COMMN SFM 0 1 $15.00 $15.00 323 0 D
# 2 COMMN 111 07/01/2009 12:00:00 AM 06/30/2010 12:00:00 AM COMMN SFM 0 1 $25.00 $25.00 323 0 D
# 3 COMMN 112 07/01/2009 12:00:00 AM 06/30/2010 12:00:00 AM COMMN SFM 0 1 $50.00 $50.00 323 0 D
# 4 COMMN 114 07/01/2009 12:00:00 AM 06/30/2010 12:00:00 AM COMMN SFM 0 1 $100.00 $100.00 323 0 M.info()可以了解数据的列信息以及每列非null的行数；
print(salary_ranges.info())# <class 'pandas.core.frame.DataFrame'>
# RangeIndex: 1356 entries, 0 to 1355
# Data columns (total 13 columns):
# # Column Non-Null Count Dtype
# --- ------ -------------- -----
# 0 SetID 1356 non-null object
# 1 Job Code 1356 non-null object
# 2 Eff Date 1356 non-null object
# 3 Sal End Date 1356 non-null object
# 4 Salary SetID 1356 non-null object
# 5 Sal Plan 1356 non-null object
# 6 Grade 1356 non-null object
# 7 Step 1356 non-null int64
# 8 Biweekly High Rate 1356 non-null object
# 9 Biweekly Low Rate 1356 non-null object
# 10 Union Code 1356 non-null int64
# 11 Extended Step 1356 non-null int64
# 12 Pay Type 1356 non-null object
# dtypes: int64(3), object(10)
# memory usage: 137.8+ KB
# None
也可以使用以下方法更快速的计算缺失值的信息；
print(salary_ranges.isnull().sum())
# SetID 0
# Job Code 0
# Eff Date 0
# Sal End Date 0
# Salary SetID 0
# Sal Plan 0
# Grade 0
# Step 0
# Biweekly High Rate 0
# Biweekly Low Rate 0
# Union Code 0
# Extended Step 0
# Pay Type 0
# dtype: int64
describe方法查看定量数据的描述性统计；Pandas认为，数据只有3个定量列：Step、Union Code和Extended Step（步进、工会代码和增强步进）。先不说步进和增强步进，很明显工会代码不是定量的。虽然这一列是数，但这些数不代表数量，只代表某个工会的代码
print( salary_ranges.describe())# Step Union Code Extended Step
# count 1356.000000 1356.000000 1356.000000
# mean 1.294985 392.676991 0.150442
# std 1.045816 338.100562 1.006734
# min 1.000000 1.000000 0.000000
# 25% 1.000000 21.000000 0.000000
# 50% 1.000000 351.000000 0.000000
# 75% 1.000000 790.000000 0.000000
# max 5.000000 990.000000 11.000000
最值得注意的特征是一个定量列Biweekly High Rate（双周最高工资）和一个定性列Grade（工作种类）；
salary_ranges = salary_ranges[['BiweeklyHighRate', 'Grade']]
print(salary_ranges.head())# BiweeklyHighRate Grade
# 0 $0.00 0
# 1 $15.00 0
# 2 $25.00 0
# 3 $50.00 0
# 4 $100.00 0
查看两个字段的类型；
print(salary_ranges.info())# <class 'pandas.core.frame.DataFrame'>
# RangeIndex: 1356 entries, 0 to 1355
# Data columns (total 2 columns):
# # Column Non-Null Count Dtype
# --- ------ -------------- -----
# 0 BiweeklyHighRate 1356 non-null object
# 1 Grade 1356 non-null object
# dtypes: object(2)
# memory usage: 21.3+ KB
# None
我们清理一下数据，移除工资前面的美元符号，保证数据类型正确。当处理定量数据时，一般使用整数或浮点数作为类型（最好使用浮点数）；定性数据则一般使用字符串或Unicode对象。
salary_ranges['BiweeklyHighRate'] = salary_ranges['BiweeklyHighRate'].map(lambda value:value.replace('$',''))
print(salary_ranges.head())# BiweeklyHighRate Grade
# 0 0.00 0
# 1 15.00 0
# 2 25.00 0
# 3 50.00 0
# 4 100.00 0
数据类型并没有变
print(salary_ranges.info())
# <class 'pandas.core.frame.DataFrame'>
# RangeIndex: 1356 entries, 0 to 1355
# Data columns (total 2 columns):
# # Column Non-Null Count Dtype
# --- ------ -------------- -----
# 0 BiweeklyHighRate 1356 non-null object
# 1 Grade 1356 non-null object
# dtypes: object(2)
# memory usage: 21.3+ KB
# None
将BiweeklyHighRate和Grade列中的数据分别转换为浮点数、字符串；
salary_ranges['BiweeklyHighRate'] = salary_ranges['BiweeklyHighRate'].astype(float)
salary_ranges['Grade'] = salary_ranges['Grade'].astype(str)
print(salary_ranges.info())# <class 'pandas.core.frame.DataFrame'>
# RangeIndex: 1356 entries, 0 to 1355
# Data columns (total 2 columns):
# # Column Non-Null Count Dtype
# --- ------ -------------- -----
# 0 BiweeklyHighRate 1356 non-null float64
# 1 Grade 1356 non-null object
# dtypes: float64(1), object(1)
# memory usage: 21.3+ KB
# None
)
