如何根据另一个变量的范围简单地创建新变量

Question

假设var1是连续的:

clear
set obs 1000
gen var1 = runiform()
sum var1

现在我想根据var1的范围创建var2。我可以这样做:

gen var2 = "Lowest" if var1<.25
replace var2 = "Low" if var1>=.25 & var1<.5
replace var2 = "High" if var1>=.5 & var1<.75
replace var2 = "Highest" if var1>=.75

我希望能够在一行中完成此操作。伪代码:

gen var2 = (ranges(0 .25 .5 .75 1) values("Lowest" "Low" "High" "Highest"))

使用cut在R中做一些非常类似的事情的方法是在R中基于范围创建分类变量

是否有任何命令可以在Stata中做一些像R版本一样的事情?想象一下，有10,000个范围需要进入var2。那么一个更好的方法会有很大帮助。

在Stata中单行执行此操作的另一种方法很笨拙，可以在http://www.stata.com/support/faqs/data-management/multiple-operations/:

找到。

generate var2 = cond(var1<=.25, "Lowest", cond(var1<=.50, "Low", cond(var1<=.75, "High", cond(var1<=1.00, "Highest", ""))))

有更好的方法吗?

Answer 1

cond()函数就是刚才提到的那个笨拙的函数。参见下面的var3示例。它具有信号优势，可以在代码中显式地显示不等式，并且完全符合您的要求，而egen, cut()都不是这样。

在这个特殊的例子中，至少还有一个技巧是可能的。参见下面的var4了解它是什么。

. clear
. set obs 15
number of observations (_N) was 0, now 15
. set seed 2803 
. gen var1 = runiform()
. sort var1 
. gen var2 = "Lowest" if var1<.25
(9 missing values generated)
. replace var2 = "Low" if var1>=.25 & var1<.5
(4 real changes made)
. replace var2 = "High" if var1>=.5 & var1<.75
(2 real changes made)
. replace var2 = "Highest" if var1>=.75
variable var2 was str6 now str7
(3 real changes made)
. gen var3 = cond(var1 < .25, "Lowest", cond(var1 <.5, "Low", cond(var1 <.75, "
> High", "Highest"))) 
. gen var4 = word("Lowest Low High Highest", ceil(4 * var1)) 
. list 
     +----------------------------------------+
     |     var1      var2      var3      var4 |
     |----------------------------------------|
  1. | .0200225    Lowest    Lowest    Lowest |
  2. | .0360774    Lowest    Lowest    Lowest |
  3. | .0934085    Lowest    Lowest    Lowest |
  4. | .0950848    Lowest    Lowest    Lowest |
  5. | .1040797    Lowest    Lowest    Lowest |
     |----------------------------------------|
  6. | .1795591    Lowest    Lowest    Lowest |
  7. | .3326341       Low       Low       Low |
  8. | .3383934       Low       Low       Low |
  9. | .3870576       Low       Low       Low |
 10. | .3980427       Low       Low       Low |
     |----------------------------------------|
 11. | .6264514      High      High      High |
 12. | .6305373      High      High      High |
 13. | .7739685   Highest   Highest   Highest |
 14. | .7935746   Highest   Highest   Highest |
 15. | .9243789   Highest   Highest   Highest |
     +----------------------------------------+

然而，如果你真的有10,000个范围要指定，并且它们不能归结为一些简单的规则，那么你自然不会使用这两种方法。您应该将它们放在一个文件中，并使用一些基于merge的代码。

Answer 2

Stata确实有一个cut函数，作为egen命令的一部分。使用它的选项并定义和分配一个值标签，可以得到想要的结果(虽然是三行而不是一行，但它们是相当简洁的三行)。。,

clear
set obs 15
gen var1 = runiform()
sum var1
gen var2 = "Lowest" if var1<.25
replace var2 = "Low" if var1>=.25 & var1<.5
replace var2 = "High" if var1>=.5 & var1<.75
replace var2 = "Highest" if var1>=.75
// =======================================================
// Using egen , cut()
// =======================================================
label define rank 0 "Lowest" 1 "Low" 2 "High" 3 "Highest"
egen var3 = cut(var1) , at(0(.25)1) icodes
label values var3 rank
li

显示结果

     +------------------------------+
     |     var1      var2      var3 |
     |------------------------------|
  1. | .6658295      High      High |
  2. | .3690664       Low       Low |
  3. | .5983131      High      High |
  4. | .2658775       Low       Low |
  5. | .1211114    Lowest    Lowest |
     |------------------------------|
  6. | .2296222    Lowest    Lowest |
  7. | .7229139      High      High |
  8. | .2501513       Low       Low |
  9. | .7775574   Highest   Highest |
 10. | .2839603       Low       Low |
     |------------------------------|
 11. | .8396428   Highest   Highest |
 12. | .4838379       Low       Low |
 13. | .2610629       Low       Low |
 14. | .3855471       Low       Low |
 15. | .3447088       Low       Low |
     +------------------------------+