累乘法计算事件收益率。  需要注意的问题：

1.数据中间有缺失（我的办法是赋值为0，还有其他办法）；

2.上市公司2千多家，如果计算所有公司的累乘（if set time interval:180/360days）计算耗用memory 较大，STATA不够用，电脑内存大应该可以克服这个问题，alternatively, 把数据分割后处理，最后合并。

参考：http://bbs.pinggu.org/thread-881526-1-1.html

code	date
1	20010102
1	20010103
1	20010105
1	20010506
2	20030201
2	20030206
2	20030406

上面的表格是事件数据

code	date	ret
1	20010101	0.001
1	20010102	0.011
1	20010103	0.021
1	20010104	0.031
1	20010105	0.041
1	20010106	0.051
1	20010107	0.061
1	20010108	0.071
1	20010109	0.081
1	20010110	0.091
1	20010505	0.101
1	20010506	0.111
1	20010507	0.121
1	20010508	0.131
1	20010509	0.141
1	20010510	0.151
1	20010511	0.161
1	20010512	0.171
1	20010513	0.181
2	20030201	0.191
2	20030202	0.201
2	20030203	0.211
2	20030204	0.221
2	20030205	0.231
2	20030206	0.241
2	20030207	0.251
2	20030405	0.261
2	20030406	0.271
2	20030407	0.281
2	20030408	0.291

上面的表格是收益率数据

code	date	rrer(10)
1	20010102
1	20010103
1	20010105
1	20010506
2	20030201
2	20030206
2	20030406

现在要计算上面的表格，也就是对每一个事件发生之后10日的累积股票收益率(累积乘法 方法计算)
注意，有的日子里有事件发生但是不是交易日，这样子的情况，需要将事件发生日期顺延到下一个交易日。

 

 

 

 

data a;

        input code date:yymmdd8.;

format date yymmdd10.;

datalines;

1 20010101

1 20010103

1 20010105

1 20010506

2 20030201

2 20030206

2 20030406

;

run;

 

data b;

        input code date:yymmdd8. ret;

        id+1;

        format date yymmdd10.;

datalines;

1 20010101 0.001

1 20010102 0.011

1 20010103 0.021

1 20010104 0.031

1 20010105 0.041

1 20010106 0.051

1 20010107 0.061

1 20010108 0.071

1 20010109 0.081

1 20010110 0.091

1 20010505 0.101

1 20010506 0.111

1 20010507 0.121

1 20010508 0.131

1 20010509 0.141

1 20010510 0.151

1 20010511 0.161

1 20010512 0.171

1 20010513 0.181

2 20030201 0.191

2 20030202 0.201

2 20030203 0.211

2 20030204 0.221

2 20030205 0.231

2 20030206 0.241

2 20030207 0.251

2 20030405 0.261

2 20030407 0.271

2 20030408 0.281

2 20030409 0.291

;

run;

 

proc sql;

        create table c as

                select

                        a.*

                        ,min(b.date) as evt_date format=yymmdd10.

                        ,min(b.id) as evt_id

                from a a

                inner join b b

                on a.code=b.code

                        and a.date<=b.date

                group by

                        a.code

                        ,a.date

        ;

quit;

 

proc sql;

        create table d as

                select

                        a.*

                        ,exp(sum(log(b.ret))) as rrer label='rrer(10)'

                from c as a

                inner join b as b

                on a.code=b.code

                        and b.id-a.evt_id between 0 and 9

                group by

                        a.code

                        ,a.date

                        ,a.evt_date

                        ,a.evt_id

                ;

quit;

STATA  例子： 

计算累计的收益率 比较简单的代码。

clear all

set mem 500m

set more off

use d:\result01

rename stkcd code

rename trddt date

replace dretwd=0 if dretwd==.

keep if code=="000001"

*drop in 1/1

*drop if date>"2014-03-05"

sort code date

gen rate=dretwd+1

capture drop product1

by code:gen product1=rate

forv i=2/180 {

local j=`i'-1

g product`i'=rate*product`j'[_n+1]

}

gen ind_return=product180-1

*keep date cdretwdtl mkt_return

save d:\kkk01.dta,replace