【stata】求滚动波动情况

0.计算对象

计算$t$、$t+1$、$t+2$ 这三起滚动波动情况
$$Vo{l}_{i,t}=ln(\sqrt{\frac{{\sum }_{n=t}^{n=t+2}(g_n-\bar{g}{)}^2}{3}})$$

$e.g$: 假设 2008-2010的value分别为【6.264004 】、 【11.11958】 、 【18.7657】
三年均值$\bar{g}$等于【12.049761】

此处数据与stata模拟数据一致，便于对比手动计算结果和代码计算结果是否一致，
进而证明代码的有效性。

. di (6.264004  + 11.11958 + 18.7657) / 3
12.049761

$$Vo{l}_{i,2008}=ln(\sqrt{\frac{{\sum }_{n=2008}^{n=2010}(g_n-\bar{g}{)}^2}{3}})$$
$$Vo{l}_{i,2008}=ln(\sqrt{\frac{(6.26-12.05{)}^2+(11.12-12.05{)}^2+(18.77-12.05{)}^2}{3}})$$
$$Vo{l}_{i,2008}=ln(\sqrt{\frac{79.5474}{3}})$$
$$Vo{l}_{i,2008}=ln(\sqrt{26.5158})$$
$$Vo{l}_{i,2008}=ln(5.1493495)$$
$$Vo{l}_{i,2008}=1.63$$

1.模拟数据

clear
set seed 123
set obs 10// ssc install egenmore,replace
egen id = repeat(),v(1/2)
sort id 
egen year = repeat(),v(2008/2012)
gen value = uniform() * 20
replace year = year - 1 if id == 2list id year value

运行结果

Number of observations (_N) was 0, now 10.(5 real changes made)+----------------------+| id   year      value ||----------------------|1. |  1   2008   6.264004 |2. |  1   2009   11.11958 |3. |  1   2010    18.7657 |4. |  1   2011   14.72644 |5. |  1   2012   3.848151 ||----------------------|6. |  2   2007   3.902801 |7. |  2   2008    19.0192 |8. |  2   2009   5.808908 |9. |  2   2010   16.38165 |10. |  2   2011   9.764193 |+----------------------+

2.根据unit id 拆分成多个dta

egen count = group(id)
su count,d
global dim = r(max)forv i = 1/$dim{preservekeep if count == `i'drop countsave temp`i'.dta,replace restore
}

3.对于每一个unit滚动求和

先定义滚动3年
然后求这3年内的value的平均 $\bar{g}$
再用逐年的数值减去均值，
而后求平方项 / 3
最后取对数

forv id = 1/$dim{
use temp`id'.dta,replace
su year ,dglobal row_year = 3  // 三年滚动求和global range_low = r(min)
global range_high = r(max) - $row_year + 1forv i = $range_low/$range_high{local j = `i' + $row_year - 1gen cond_bin = (year >= `i' & year <= `j')egen m_`i' = mean(value) if cond_bin == 1ereplace m_`i' = mean(m_`i')gen rolling_`i' = (value - m_`i')^2 if cond_bin == 1ereplace rolling_`i' = total(rolling_`i') if cond_bin == 1replace rolling_`i' = ln(sqrt(rolling_`i' / $row_year))drop cond_bin
}
keep id rolling*
forv i = $range_low/$range_high{ereplace rolling_`i' = mean(rolling_`i')
}
duplicates drop id ,force 
reshape long rolling_,i(id) j(year)
save rolling_`id'.dta,replace
}

4.合并/清除数据

use rolling_1.dta,replace
forv id = 2/$dim{append using rolling_`id'.dta
}
forv i = 1/$dim{erase temp`i'.dtaerase rolling_`i'.dta
}
save rolling.dta,replace 

5.运行结果

use rolling.dta,clear
list id year roll in 1/6

结果对比：可以看到，id=1时，year=2008时，代码计算结果也为1.63，与手动计算结果一致，证明了代码的有效性

     +----------------------+| id   year   rolling_ ||----------------------|1. |  1   2008   1.638221 |2. |  1   2009   1.138851 |3. |  1   2010   1.840516 |4. |  2   2007   1.905365 |5. |  2   2008   1.741902 |6. |  2   2009   1.472856 |+----------------------+