Question

I have a dataset in wide format that has quarterly counts of an event from Q1 1996 to Q4 2016.

The variable names for each quarter are as follows:

Q1 = yyyy0101_yyyy0401
Q2 = yyyy0401_yyyy0701
Q3 = yyyy0701_yyyy1001
Q4 = yyyy1001_yyyy0101

I have a macro that converts them like this:

local i = 1996

forvalues x = 1996/2016 { 
    local i = `i'+1
    gen count`x' = event_`x'0101_`x'0401 + event_`x'0401_`x'0701 + 
    event_`x'0701_`x'1001 + event_`x'1001_`i'0101
}

Then I collapse my data into a single variable count in long format by year:

reshape long count, i(xvars) j(year)

Now I would like to do the same thing, but quarterly.

What is the macro to perform the exact same process, but capturing the sum of counts by year-quarter? What if I wanted to do it for half years?

Answer 1

在试图理解这一点时，我首先（迂腐地或其他方面）注意到你的代码不是Stata意义上的宏：SAS用户???）和第二个（更具建设性）你的代码可以被压缩到

forvalues x = 1996/2016 { 
    gen count`x' = event_`x'0101_`x'0401 + event_`x'0401_`x'0701 + event_`x'0701_`x'1001 + event_`x'1001_`x'0101
}

鉴于本地宏i和x运行相同的值。（这些都是Stata意义上的宏。）但我不会从那里开始。

正如你所说，数据是宽格式的（我更喜欢术语布局减少重载），所以主要工作不是写循环，而是reshape长。这个例子体现了一些猜测并展示了一些技巧。如果没有要使用的数据示例，我首先创建一个沙箱：

clear 
set obs 21
local y = 1 
foreach v in yyyy0101_yyyy0401 yyyy0401_yyyy0701 yyyy0701_yyyy1001 yyyy1001_yyyy0101 { 
    gen `v' = `++y' 
}
gen year = 1995 + _n 

rename (yyyy*) (count#), addnumber 

reshape long count, i(year) j(quarter) 

gen qdate = yq(y, q) 

egen ycount = total(count), by(year) 
egen qcount = total(count), by(quarter) 
gen half = cond(inlist(quarter, 1, 2), 1, 2)
egen hcount = total(count), by(year half) 

list if year < 1998, sepby(year) 

     +------------------------------------------------------------------+
     | year   quarter   count   qdate   ycount   qcount   half   hcount |
     |------------------------------------------------------------------|
  1. | 1996         1       2     144       14       42      1        5 |
  2. | 1996         2       3     145       14       63      1        5 |
  3. | 1996         3       4     146       14       84      2        9 |
  4. | 1996         4       5     147       14      105      2        9 |
     |------------------------------------------------------------------|
  5. | 1997         1       2     148       14       42      1        5 |
  6. | 1997         2       3     149       14       63      1        5 |
  7. | 1997         3       4     150       14       84      2        9 |
  8. | 1997         4       5     151       14      105      2        9 |
     +------------------------------------------------------------------+

与您使用reshape long时的情况有何不同？主要是要强调在不同时间尺度上你想要的任何总数都可以在适当的位置生成。您不需要重复collapse个或同一数据集的不同版本。如何制表，列表或以其他方式处理重复将是不同的问题。

Year to month date conversion

1 个答案: