CHP03- Data , Variable & Label

1. 查看数据

sysuse auto, clear
browse           *浏览数据
list make price mpg in 1/20  *列出数据
describe         *描述数据
describe, detail
d make price mpg
summarize        *基本统计量
summarize,detail
sum make price mpg
codebook         *列出变量信息
inspect          *描述变量的属性

2. 数据类型( storage type )

2.1. numeric

                                                         Closest to
    Storage                                              0 without
    type                 Minimum              Maximum    being 0     bytes
    ----------------------------------------------------------------------
    byte                    -127                  100    +/-1          1
    int                  -32,767               32,740    +/-1          2
    long          -2,147,483,647        2,147,483,620    +/-1          4
    float   -1.70141173319*10^38  1.70141173319*10^38    +/-10^-38     4
    double  -8.9884656743*10^307  8.9884656743*10^307    +/-10^-323    8
    ----------------------------------------------------------------------
    Precision for float  is 3.795x10^-8.
    Precision for double is 1.414x10^-16.

2.2 string

    String
    storage       Maximum
    type          length         Bytes
    -----------------------------------------
     str1             1             1
     str2             2             2
      ...             .             .
      ...             .             .
      ...             .             .
     str2045         2045           2045

     strL            2000000000     2000000000
    -----------------------------------------

2.3 datetime

2.3.1 Types of dates–human readable forms (HRFs)

方便阅读，为字符型

    Date type         Examples of HRFs
    --------------------------------------------
    datetime          20jan2010 09:15:22.120  
       
    date              20jan2010, 20/01/2010, ...
    
    weekly date       2010w3
    monthly date      2010m1
    quarterly date    2010q1
    half-yearly date  2010h1
    yearly date       2010
    --------------------------------------------

2.3.2 Types of dates–Stata internal form (SIF)

方便计算及提取，为数值型

    SIF type        Examples in SIF       Units
    -----------------------------------------------------------------
    datetime/c      1,579,598,122,120     milliseconds since 
                                          01jan1960 00:00:00.000, 
                                          assuming 86,400 s/day
    
    datetime/C      1,579,598,146,120     milliseconds since 
                                          01jan1960 00:00:00.000, 
                                          adjusted for leap seconds*
                                          
    date                       18,282     days since 01jan1960
                                          (01jan1960 = 0)
                                          
    weekly date                 2,601     weeks since 1960w1
    monthly date                  600     months since 1960m1
    quarterly date                200     quarters since 1960q1
    half-yearly date              100     half-years since 1960h1
    yearly date                  2010     years since 0000
    -----------------------------------------------------------------
    SIF datetime/C is equivalent to coordinated universal time (UTC). 
    In UTC, leap seconds are periodically inserted because the length of the mean solar day is slowly increasing.

2.3.3 HRF-to-SIF conversion functions

                         Function to convert
    SIF type        HRF to SIF                     Note
    --------------------------------------------------------------------
    datetime/c      tc =      clock(HRFstr, mask)  tc must be double
    datetime/C      tC =      Clock(HRFstr, mask)  tC must be double
    
    date            td =       date(HRFstr, mask)  td may be float or
                                                              long
    
    weekly date     tw =     weekly(HRFstr, mask)  tw may be float or int
    monthly date    tm =    monthly(HRFstr, mask)  tm may be float or int
    quarterly date  tq =  quarterly(HRFstr, mask)  tq may be float or int
    half-year date  th = halfyearly(HRFstr, mask)  th may be float or int
    yearly date     ty =     yearly(HRFstr, mask)  ty may be float or int
    --------------------------------------------------------------------
    Warning: To prevent loss of precision, datetime SIFs must be stored as doubles.

examples：
- 字符型变量mystr，值如"2010.07.12 14:32"或"2010 Jul 12 14:32"或"July 12, 2010 2:32 PM"，转为数值型命令为(stata会自动识别相应时间的表达形式)
  gen double eventtime = clock(mystr, "YMDhm")
- 值如"2010.07.12 14:32:12"，转为数值型命令为
  gen double eventtime = clock(mystr, "YMDhms")
- 值如"7-12-10 14.32"，2位数的年份需要给定年份前缀，如20xx年
  gen double eventtime = clock(mystr, "MD20Yhm")
- 值如"14:32 on 7/12/2010"，需用#忽略其中的多余字符
  gen double eventtime = clock(mystr, "hm#MDY")
- 值如"22/7/2010"，以下均可
  gen eventdate = date(mystr, "DMY")
  gen double eventtime = clock(mystr, "DMY")

2.3.4 Displaying SIFs in HRF

SIF格式的日期数据，本质上是一串数字，代表距离1960年1月1日的微秒数。
为使SIF也像HRF那样方便阅读，可以设定其显示格式
注意：显示成HRF，但是本质仍是数值。一个简单的区分，HRF在数据浏览窗口显示为棕色字符2010.07.12 14:32，SIF的HRF表达化，则显示为蓝色字符2010.07.12 14:32，类似于添加了label value。

                     Display format to 
    SIF type         present SIF in HRF
    -----------------------------------
    datetime/c            %tc
    datetime/C            %tC
    date                  %td
    weekly date           %tw
    monthly date          %tm
    quarterly date        %tq
    half-yearly date      %th
    yearly date           %ty
    -----------------------------------

examples：

已有字符型变量mystr，值如"2010.07.12 14:32"

gen double eventtime = clock(mystr, "YMDhm")
format eventtime %tc

已有字符型变量mystr，值如"22/7/2010"

gen eventdate = date(mystr, "DMY")
format eventdate %td

可以呈现多种效果

format eventdate %tdCY.N.D
format eventdate %tdCY/N/D
format eventdate %tdCY-M-D
format eventdate %tdCY_M_D //"_"表示空格

2.3.5 SIF-to-SIF conversion

                | To:
    From:       |     datetime/c   datetime/C   date
    ------------+------------------------------------------
    datetime/c  |                  tC=Cofc(tc)  td=dofc(tc)
    datetime/C  |     tc=cofC(tC)               td=dofC(tC)
    date        |     tc=cofd(td)  tC=Cofd(td)  
    weekly      |                               td=dofw(tw)
    monthly     |                               td=dofm(tm)
    quarterly   |                               td=dofq(tq)
    half-yearly |                               td=dofh(th)
    yearly      |                               td=dofy(ty)
    -------------------------------------------------------

                | To:
    From:       |     weekly       monthly      quarterly 
    ------------+------------------------------------------
    date        |     tw=wofd(td)  tm=mofd(td)  tq=qofd(td)
    -------------------------------------------------------

                | To:
    From:       |     half-yearly  yearly
    ------------+------------------------------------------
    date        |     th=hofd(td)  ty=yofd(td)
    -------------------------------------------------------

2.3.6 datetime数据其他整理

help datetime
- Building SIFs from components
- Extracting time-of-day components from SIFs
- Extracting date components from SIFs
- Conveniently typing SIF values
- Obtaining and working with durations
- Using dates and times from other software

2.3.7 datetime 小练习

从年月日字符串分别提取数值型年月日

generate double timestamp = date(varname, "DMY")
gen  year=year(timestamp)
gen  month=month(timestamp)
gen  day=day(timestamp)

2.4 missing values

all numbers < . < .a < .b < · · · < .z
string缺失：""，区别于 " "
missing函数：!missing(varname)会将所有缺失值排除掉，毕竟缺失值的编码有27种之多。

2.4 数据压缩

compress
- 不会改变数据，却能缩小数据占用的空间
- 通过转换数据类型来压缩数据
  
  数据类型转换后的类型
  
  double long , int 或 byte
  
  float int 或 byte
  
  long int 或 byte
  
  int byte
  
  string 短 string

数据类型	转换后的类型
double	long , int 或 byte
float	int 或 byte
long	int 或 byte
int	byte
string	短 string

2.5 更改数据存储类型

recast

sysuse auto,clear
list gear_ratio in 1/5
d gear_ratio
recast int gear_ratio, force
d gear_ratio
list gear_ratio in 1/5

3. 数据的显示格式

数值型数据格式设定 %[ - | ~ ][ 0 ] w . d [ e | f | g ][ c ]

% 表明开始设定格式
负号 "-" 表示数据靠左列式
数字 "0" 表示输入的数字 $0$ 则会保留在开头
数字 $w$ 表示显示格式的宽度
"." 为小数点
数字 $d$ 表示小数点后保留的小数位数
字母 $e$ 表示用科学计数法表示
字母 $f$ 表示固定格式
字母 $g$ 表示通用设定，即Stata会自动设定显示格式
字母 $c$ 表示将数据设置成有逗号的西方计数方法
例如：
- str18 文字型变量，每个观察值占据18个空格
- %-18s 靠左列印于屏幕上；若%18s，则靠右列印；若 %~18s, 则居中列印
- %8.0g 在 ‘8.0’ 的原则下，以尽量多的有效位数列出
- %6.2f 总共占个空格，小数位占两个空格

example

display

.dis  %7.2f  1000.12345
1000.12
.dis  %9.2e  1000.12345
1.00e+03
.dis  %8.2fc  1000.12345
1,000.12

format

. sysuse auto.dta, clear
. list price gear in 1/3
     +------------------+
     | price   gear_r~o |
     |------------------|
  1. | 4,099       3.58 |
  2. | 4,749       2.53 |
  3. | 3,799       3.08 |
     +------------------+
. format price %7.1fc   / /保留一位小数，并保留价格中的逗号
. format gear %6.3f     / /保留三位小数
. list price gear in 1/3
     +--------------------+
     |   price   gear_r~o |
     |--------------------|
  1. | 4,099.0      3.580 |
  2. | 4,749.0      2.530 |
  3. | 3,799.0      3.080 |
     +--------------------+

描述性统计

. tabstat price mpg rep78 length turn foreign, ///
>         stats(n mean sd min max) c(s) f(%9.2f)

    variable |         N      mean        sd       min       max
-------------+--------------------------------------------------
       price |     74.00   6165.26   2949.50   3291.00  15906.00
         mpg |     74.00     21.30      5.79     12.00     41.00
       rep78 |     69.00      3.41      0.99      1.00      5.00
      length |     74.00    187.93     22.27    142.00    233.00
        turn |     74.00     39.65      4.40     31.00     51.00
     foreign |     74.00      0.30      0.46      0.00      1.00
----------------------------------------------------------------

估计结果的显示格式

. qui reg price mpg
. est store reg1
. qui reg price mpg rep78
. est store reg2
* 输出估计结果
. esttab reg1 reg2, b(%9.2f) se(%7.2f) sfmt(%7.3f) / / /
>         star(* 0.1 ** 0.05 *** 0.01) mtitle(reg1 reg2) / / /
>         scalar(r2 r2_a N F) compress nogap

------------------------------------
                 (1)          (2)   
                reg1         reg2   
------------------------------------
mpg          -238.89***   -271.64***
             (53.08)      (57.77)   
rep78                      666.96*  
                         (342.36)   
_cons       11253.06***   9657.75***
           (1170.81)    (1346.54)   
------------------------------------
N                 74           69   
r2             0.220        0.251   
r2_a           0.209        0.228   
F             20.258       11.057   
------------------------------------
Standard errors in parentheses
* p<0.1, ** p<0.05, *** p<0.01

图片中数字的显示格式

#delimit ;
twoway (scatter price mpg) (lfit price mpg),
    ylabel(, format(%9.1f) angle(0) nogrid)
    xlabel(, format(%3.1f))
;
#delimit cr

4. 变量

4.1 变量生成

generate & replace

*Syntax
generate [type] newvar[:lblname] =exp [if] [in] [, before(varname) | after(varname)]
replace oldvar =exp [if] [in] [, nopromote]

*Examples
webuse genxmpl3,clear
generate age2 = age^2
generate int age2 = age^2
webuse genxmpl1, clear
replace age2 = age^2

4.2 变量命名

由英为字母、数字或 _ 组成，至多不超过 32 个；
首字母必须为字母或 _ ；
英文字母的大写和小写具有不同的含义；
- 例如：abc_1 a1 _a2 Gdp 都是合理的变量名
  5gdp 2invest 则不是；
建议不要使用 _ 作为变量的第一个字母，
- 因为许多stata的内部变量都是以 _ 开头的，
- 如，_n, _N, _cons, _b 等等。
变量重命名
- rename old_varname new_varname
- example
```
sysuse auto,clear 
d
rename rep78 repair78
d
```

4.3 变量类型

区别于数据存储类型
- 数据存储类型分为数值型与字符型（存储角度）
- 变量类型分为定类变量与定性变量（分析角度）

graph LR A((Variable)) --- B((Quantitative Variable)); A((Variable)) --- C((Qualitative Variable)); B((Quantitative Variable)) --- D(Continuous Variable); B((Quantitative Variable)) --- E(Discrete Variable); C((Qualitative Variable)) --- F(Binary Variable); C((Qualitative Variable)) --- G(Nominal Variable); C((Qualitative Variable)) --- H(Ordinal Variable);

5. 标签 ( label )

5.1 数据标签

*Syntax
label data ["label"]
*Examples
sysuse auto,clear
d
label data "1978年汽车价格资料数据"
d   //注意观察数据标签

5.2 变量标签

*Syntax
label variable varname ["label"]
*Examples
sysuse auto,clear
label var price    汽车价格
label var foreign  "汽车产地(1 国外; 2 国内)"

5.3 值标签

*Syntax
* Define value label
label define lblname # "label" [# "label" ...] [, add modify replace nofix]
* Assign value label to variables
label values varlist lblname [, nofix]
* Remove value labels
label values varlist [.]
* List names of value labels
label dir
* List names and contents of value labels
label list [lblname [lblname ...]]
* Copy value label
label copy lblname lblname [, replace]
* Drop value labels
label drop {lblname [lblname ...] | _all}
* Save value labels in do-file
label save [lblname [lblname...]] using filename [, replace]

*Examples
sysuse auto,clear
* label define 标签名
* label values 变量名 标签名 /*将变量值和标签联系起来*/
label define repair  1 "好" 2 "较好" 3 "中" 4 "较差" 5 "差"
label values rep78 repair
*显示值标签
label dir
label list
label list repair
*添加和修改值标签
label def repair 5 "差", add
label def repair 3 "一般", modify
*删除值标签
label drop repair
label list

本节命令：
browse , list , describe , summarize , codebook , inspect
compress , recast , format , display , recast
gen , replace , rename , label