Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

Find a Community

- Home
- /
- SAS Programming
- /
- General Programming
- /
- Math with Observations from different Variables/Da...

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

07-15-2013 01:11 PM

Hi,

I have two Data Sets (A and B). Data Set A has 1 variable (called A1) and Data Set B has also 1 variable (called B1). Both have many observations.

I want to create a new Data Set, named C and with one variable called C1. This C1 variable should be generate as follow:

**Logic:**

A1 | B1 | C1 |

Observation_1 | Observation_1 | Absolute value of { [[A1 (Observation_1)] - [B1 (Observation_2)]} |

Observation_2 | Observation_2 | Absolute value of { [A1 (Observation_2)] - [B1 (Observation_3)]} |

Observation_3 | Observation_3 | Absolute value of { [A1 (Observation_3)] - [B1 (Observation_4)]} |

Observation_4 | Observation_4 | Absolute value of { [A1 (Observation_4)] - [B1 (Observation_5)]} |

Observation_5 | Observation_5 | Absolute value of { [A1 (Observation_5)] - [B1 (Observation_6)]} |

Observation_6 | Observation_6 | . |

**Exemple:**

A1 | B1 | C1 |

3 | 11 | 0 |

-7 | 3 | 4 |

4 | -3 | 6 |

-2 | -2 | 3 |

3 | -5 | 2 |

13 | 1 | . |

How should I proceed?

Thanks!

Accepted Solutions

Solution

07-15-2013
01:43 PM

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

07-15-2013 01:43 PM

Here's a one-step alternative:

data want;

if set_is_done=0 then set b (firstobs=2 keep=b1 rename=(b1=next_b1)) end=set_is_done;

drop next_b1;

merge a b end=merge_is_done;

if merge_is_done=0 then c1 = abs(a1 - next_b1);

run;

If you don't want to drop NEXT_B1, you would also need to reset its value to missing on the final observation. You're also responsible for being sure that the data sets contain the same number of observations when you begin. Good luck.

All Replies

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

07-15-2013 01:26 PM

Create a count variable on each dataset and then merge with n=n+1;

data a_count;

set a;

count=_n_;

run;

data b_count;

set b;

count=_n_;

run;

proc sql;

create table want as

select a.a1, b.b1, abs(a1-b1) as c1

from a_count as a

join b_count as b

on a.count=b.count+1;

quit;

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

07-15-2013 01:43 PM

You beat me to the post ... all well here is my version:

data a;

infile cards dsd;

input A1;

cards;

3

-7

4

-2

3

13

run;

data b;

infile cards dsd;

input B1;

cards;

11

3

-3

-2

-5

1

run;

data a2;

set a;

Obs= _n_ ;

run;

data b2;

set b;

Obs= _n_ ;

run;

proc sql;

create table out as

select a.a1, a.b1,

abs(a.a1 - c.b1) as c1

from (select a.obs, a.a1, b.b1 from a2 as a , b2 as b where a.obs = b.obs) as a left join b2 as c

on a.obs+1 = c.obs

;

quit;

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

07-15-2013 01:48 PM

Just a little point of correction ... I think Reeza's on statement should read on a.count+1=b.count;

EJ

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

07-15-2013 01:57 PM

I think mine is correct, but I'm too lazy to check, I'll leave it up to the OP to use what's best for them

Absolute value of { [[A1 (Observation_1)] - [B1 (Observation_2)]}

However I don't think the data would be correct, because they're asking for obs1 and obs1 and then obs1-obs2 in the dataset. Which seems weird, so I'll wait for confirmation for what the OP wants.

Solution

07-15-2013
01:43 PM

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

07-15-2013 01:43 PM

Here's a one-step alternative:

data want;

if set_is_done=0 then set b (firstobs=2 keep=b1 rename=(b1=next_b1)) end=set_is_done;

drop next_b1;

merge a b end=merge_is_done;

if merge_is_done=0 then c1 = abs(a1 - next_b1);

run;

If you don't want to drop NEXT_B1, you would also need to reset its value to missing on the final observation. You're also responsible for being sure that the data sets contain the same number of observations when you begin. Good luck.

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

07-15-2013 02:56 PM

And a similar data step alternative. SQL seems unsuited to this task.

set a;

if not eof then set b(firstobs=

C1 = abs(a1-b1);

set b;

output;

call missing(of _all_);

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

07-15-2013 04:33 PM

Thank you all guys!

Just one more question:

What if instead of "Absolute value of { [[* A1* (Observation_1)] - [

the logic would be "Absolute value of { [[* A1* (Observation_1)] - [

?

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

07-15-2013 05:07 PM

Well, the change would be relatively straightforward. To bring in the next A observation instead of the next B observation:

if set_is_done=0 then set a (firstobs=2 keep=a1 rename=(a1=next_a1)) end=set_is_done;

There would be a small change to the formula to compute c1.

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

07-15-2013 08:43 PM

I would use DIF function for that similar to LAG but with difference. The difference would be on a different observation but that shouldnt' matter.

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

07-16-2013 10:10 AM

Thanks data_null_!

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

07-15-2013 11:02 PM

Not tested code :

data want;

merge a a(firstobs=2 rename=(a1=_a1));

C1 = abs(a1-_a1);

.......

Ksharp

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

07-16-2013 10:11 AM

Thanks you too Ksharp!