HOW TO DEAL WITH THIS ONE (SURVEYSELECT)

I HAVE AN ORIGINAL DATA SET AS:
ID
1_A
1_B
1_C
1_D
2_A
2_B
2_C
2_D
.....
.....
100_A
100_B
100_C
100_D

BUT I NEED ONLY HALF OF THE DATA SET (RANDOMLY SELECTED), BUT THE PROBLEM IS THAT, ONCE 1_A WAS SELECTED, ALL 1_A, 1_B, 1_C AND 1_D SHOULD ALSO BE SELECTED.

HOW CAN I DEAL WITH THIS KIND OF PROBLEM?

THANKS A LOT!!

rockfido 发表于 2010-3-2 00:14
I HAVE AN ORIGINAL DATA SET AS:
ID
1_A
1_B
1_C
1_D
2_A
2_B
2_C
2_D
.....
.....
100_A
100_B
100_C
100_D

BUT I NEED ONLY HALF OF THE DATA SET (RANDOMLY SELECTED), BUT THE PROBLEM IS THAT, ONCE 1_A WAS SELECTED, ALL 1_A, 1_B, 1_C AND 1_D SHOULD ALSO BE SELECTED.

HOW CAN I DEAL WITH THIS KIND OF PROBLEM?

THANKS A LOT!!

The problem is equivalent to,

select numbers from [1,2,3,...,100] randomly. Once you have the numbers, you use them to subset your original data.

HTH

问2个问题：
1.你打算运行随机选择几次？是100次还是400次？
2.1_A,1_B,1_C,1_D是不是作为一个整体，要么都在要么都不在？

bobguy 发表于 2010-3-2 08:01

rockfido 发表于 2010-3-2 00:14
I HAVE AN ORIGINAL DATA SET AS:
ID
1_A
1_B
1_C
1_D
2_A
2_B
2_C
2_D
.....
.....
100_A
100_B
100_C
100_D

BUT I NEED ONLY HALF OF THE DATA SET (RANDOMLY SELECTED), BUT THE PROBLEM IS THAT, ONCE 1_A WAS SELECTED, ALL 1_A, 1_B, 1_C AND 1_D SHOULD ALSO BE SELECTED.

HOW CAN I DEAL WITH THIS KIND OF PROBLEM?

THANKS A LOT!!

The problem is equivalent to,

select numbers from [1,2,3,...,100] randomly. Once you have the numbers, you use them to subset your original data.

HTH

正解......
把scan(id,1,"_")相同的视为一个变量(组)

不错 学习了 楼上都是高人

thanks a lot!!


2# bobguy

1. /*TST*/
   
2. data tst;
   
3. do x=rank("A") to rank("Z");
   
4. do y=1 to 100 by 1;
   
5. ID=compress(y||"_"||byte(x));
   
6. output;
   
7. end;
   
8. end;
   
9. keep ID;
   
10. run;
    
11. 
    
12. /*GRP*/
    
13. data p_1;
    
14. set tst;
    
15. Grp=input(scan(ID,1,"_"),4.);
    
16. run;
    
17. 
    
18. /*SORT*/
    
19. proc sort data=p_1 out=p_2;by Grp;run;quit;
    
20. 
    
21. /*RTO of SURVEYSELECT*/
    
22. %let rto=0.2;
    
23. 
    
24. /*RST*/
    
25. data p_3;
    
26. retain r;
    
27. set p_2;
    
28. by Grp;
    
29. if first.Grp then do;
    
30. r=ranuni(1);
    
31. end;
    
32. if r<=%sysevalf(&rto);
    
33. 
    
34. run;

复制代码

没仔细测试，看看好像行
希望有用......

1.没太明白，可能我还不是太明白SAS中的随机过程，我就只会用PROC SURVEYSELECT。。。。反正我就是想选出一半数量的DATA SET。
2.是的，要么都在，要么都不在。


3# frackdeng

非常感谢，我也想给你送20个金币，可是不知道怎么送。。。。。

谁能告诉我一声？

BTW,有朋友给我另一个建议，觉得也不错，分享给大家：

可以把数字跟字母分开成两个变量，然后进行TRANSPOSE，然后再对数字变量进行RANDOM SAMPLING，然后再TRANSPOSE回来，应该也可以的。


7# soporaeternus

7楼的很好用 学习了