Given two simple datasets A and B as follows
DATA A; INPUT X ##;
CARDS;
1 2 3 4
RUN;
DATA B; INPUT Y ##;
CARDS;
1 2
RUN;
I am trying to create two datasets named C and D, one using repeated SET and OUTPUT statements and another using DO loop.
DATA C;
SET B;
K=1; DO; SET A; OUTPUT; END;
K=K+1; DO; SET A; OUTPUT; END;
K=K+1;
RUN;
DATA D;
SET B;
DO K = 1 TO 2;
SET A; OUTPUT;
END;
RUN;
I thought that C and D should be the same as the DO loop is supposed to be repeating those statements as shown in the DATA step for C, but it turns out that they are different.
Dataset C:
Obs Y K X
1 1 1 1
2 1 2 1
3 2 1 2
4 2 2 2
Dataset D:
Obs Y K X
1 1 1 1
2 1 2 2
3 2 1 3
4 2 2 4
Could someone please explain this?
The two SET A statements in the first data step are independent. So on each iteration of the data step they will both read the same observation. So it is as if you ran this step instead.
data c;
set b;
set a;
do k=1 to 2; output; end;
run;
The SET A statement in the second data step will execute twice on the first iteration of the data step. So it will read two observations from A for each iteration of the data step.
If you really wanted to do a cross-join you would need to use point= option so that you could re-read one of the data sets.
data want ;
set b ;
do p=1 to nobs ;
set a point=p nobs=nobs ;
output;
end;
run;
Your Table B has two obs so your code will only do two iterations:
Every time you read a new observation K resets to 1, Solution: use Retain keyword.
When your current records is OBS 1 and you do an output, you will keep outputting the first row from each table, that's why you output the first and second rows twice from table A.
Debugging:
Iteration 1 current view:
Obs Table X
1 A 1
Obs Table Y k
1 B 1 1
Output:
K=1; DO; SET A; OUTPUT; END;
Obs Y K X
1 1 1 1
K=K+1; DO; SET A; OUTPUT; END;
Obs Y K X
2 1 2 1
Iteration 2 current view:
Obs Table X
2 A 2
Obs Table Y k
2 B 2 1
Output:
K=1; DO; SET A; OUTPUT; END;
Obs Y K X
3 2 1 2
K=K+1; DO; SET A; OUTPUT; END;
Obs Y K X
4 2 2 2
Related
I have the below requirement and wondering if this can be implemented in simple datastep:
will be starting with a simple dataset with two variable
input:
x y
1 0
logic:
x y z
1 0 x+y
prev z prev y +1 x+y
prev z prev y +1 x+y
output:
x y z
1 0 1
1 1 2
2 2 4
4 3 7
7 4 11
Just output the computed rows one-by-one within a do-while loop.
data have;
input x y;
cards;
1 0
;run;
data want(drop=i);
set have;
z = x + y;
output;
i = 2; /*next row*/
do while (i <= 5); /* put the total number of rows here */
x = z;
y = y + 1;
z = x + y;
output;
i = i + 1;
end;
run;
Result
proc print data=want;
run;
Obs x y z
1 1 0 1
2 1 1 2
3 2 2 4
4 4 3 7
5 7 4 11
Macro version:
%macro gen(have, want, n_rows);
data &want(drop=i);
set &have;
z = x + y;
output;
i = 2;
do while (i <= &n_rows);
x = z;
y = y + 1;
z = x + y;
output;
i = i + 1;
end;
run;
%mend gen;
/* execute */
%gen(have, want, 5)
Ok. I decided to do some research to try to answer your question. And I found out that you (probably) did the same question in a sas forum (link here).
As any interested user may see, the question was answered in a very elegant way there (by Mr. Reinhard - credits to him).
While I was pasting Reinhard answer here, I saw that #Bill Huang came with an original answer his own. So probably you should accept his answer. Anyway, Reinhard answer was really cool and elegant, so I thought it might worth to have it registerd here. Mainly for other users, because it is such an easy way of creating additional interative rows in SAS:
data have;
x=1; y=0;
run;
data want;
set have;
do y=y to 4;
z=x+y;
output;
x=z;
end;
#LuizZ's do y=y to 4 version presumes the first row has y=0. If that is not the case try
data have;
x=1; y=0;
run;
%let iterations = 4;
data want;
set have;
do y = y to y+&iterations;
z = x + y;
output;
x = z;
end;
run;
In this data, I need to subset by each variable by certain percentage.
For example,
Obs Group Score
1 A 1
2 A 2
3 B 1
4 B 1
5 C 3
6 C 1
7 C 1
8 A 1
9 A 3
10 A 1
11 A 2
12 B 3
13 C 2
I would need to subset 10 obs.
The sample must consist of all groups, and score of 1 takes higher priority.
Each group is given certain percent.
Let say 50% for A, 20% for B and 30% for C.
I tried using proc surveyselect but it failed. The number of alloc is not same as the strata.
proc surveyselect data=example out=test sampsize=10;
strata group score/alloc=(0.5 0.2 0.3);
run;
I don't know proc surveyselect too much, so I give the data step version.
data have;
input Obs Group$ Score;
cards;
1 A 1
2 A 2
3 B 1
4 B 1
5 C 3
6 C 1
7 C 1
8 A 1
9 A 3
10 A 1
11 A 2
12 B 3
13 C 2
;
run;
proc sort;
by Group Score;
run;
data want;
array _Dist_[3]$ _temporary_('A','B','C');
array _Upper_[3] _temporary_(5,2,3);
array _Count_[3] _temporary_;
do i = 1 to rec;
set have nobs=rec point=i;
do j = 1 to dim(_Dist_);
_Count_[j] + (Group=_Dist_[j]);
if _Count_[j] <= _Upper_[j] and Group = _Dist_[j] then output;
end;
end;
stop;
drop j;
run;
I have a row matrix (vector) A and another square matrix B. How can I multiply each row of matrix B with the row matrix A in SAS using proc iml or otherwise?
Let's say
a = {1 2 3}
b =
{2 3 4
1 5 3
5 9 10}
My output c would be:
{2 6 12
1 10 9
5 18 30}
Thanks!
Use the element-wise multiplication operator, # in IML:
proc iml;
a = {1 2 3};
b = {2 3 4,
1 5 3,
5 9 10};
c = a#b;
print c;
quit;
There's of course a non-IML solution, or twenty, though IML as Dom notes is probably easiest. Here's two.
First, get them onto one dataset, where the a dataset is on every row (with some other variable names) - see below. Then, either just do the math (use arrays) or use PROC MEANS or similar to use the a dataset as weights.
data a;
input w_x w_y w_z;
datalines;
1 2 3
;;;;
run;
data b;
input x y z;
id=_n_;
datalines;
2 3 4
1 5 3
5 9 10
;;;;
run;
data b_a;
if _n_=1 then set a;
set b;
*you could just multiply things here if you wanted;
run;
proc means data=b_a;
class id;
types id;
var x/weight=w_x;
var y/weight=w_y;
var z/weight=w_z;
output out=want sum=;
run;
I answered a SAS question a few minutes ago and realized there is a generalization that might be more useful than that one (here). I didn't see this question already in StackOverflow.
The general question is: How can you process and keep an entire BY-group based on some characteristic of the BY-group that you might not know until you have looked at all the observations in the group?
Using input data similar to that from the earlier question:
* For some reason, we are tasked with keeping only observations that
* are in groups of ID_1 and ID_2 that contain at least one obs with
* a VALUE of 0.;
* In the following data, the following ID and ID_2 groups should be
* kept:
* A 2 (2 obs)
* B 1 (3 obs)
* B 3 (2 obs)
* B 4 (1 obs)
* The resulting dataset will have 8 observations.;
data x;
input id $ id_2 value;
datalines;
A 1 1
A 1 1
A 1 1
A 2 0
A 2 1
B 1 0
B 1 1
B 1 3
B 2 1
B 3 0
B 3 0
B 4 0
C 2 4
;
run;
Double DoW loop solution:
data have;
input id $ id_2 value;
datalines;
A 1 1
A 1 1
A 1 1
A 2 0
A 2 1
B 1 0
B 1 1
B 1 3
B 2 1
B 3 0
B 3 0
B 4 0
C 2 4
;
run;
data want;
do _n_ = 1 by 1 until(last.id_2);
set have;
by id id_2;
flag = sum(flag,value=0);
end;
do _n_ = 1 to _n_;
set have;
if flag then output;
end;
drop flag;
run;
I've tested this against the point approach using ~55m rows and found no appreciable difference in performance. Dataset used:
data have;
do ID = 1 to 10000000;
do id_2 = 1 to ceil(ranuni(1)*10);
do value = floor(ranuni(2) * 5);
output;
end;
end;
end;
run;
My answer might not be the most efficient, especially for large datasets, and I'm interested in seeing other possible answers. Here it is:
* For some reason, we are tasked with keeping only observations that
* are in groups of ID_1 and ID_2 that contain at least one obs with
* a VALUE of 0.;
* In the following data, the following ID and ID_2 groups should be
* kept:
* A 2 (2 obs)
* B 1 (3 obs)
* B 3 (2 obs)
* B 4 (1 obs)
* The resulting dataset will have 8 observations.;
data x;
input id $ id_2 value;
datalines;
A 1 1
A 1 1
A 1 1
A 2 0
A 2 1
B 1 0
B 1 1
B 1 3
B 2 1
B 3 0
B 3 0
B 4 0
C 2 4
;
run;
* I realize the data are already sorted, but I think it is better
* not to assume they are.;
proc sort data=x;
by id id_2;
run;
data obstokeep;
keep id id_2 value;
retain startptr haszero;
* This SET statement reads through the dataset in sequence and
* uses the CUROBS option to obtain the observation number. In
* most situations, this will be the same as the _N_ automatic
* variable, but CUROBS is probably safer.;
set x curobs=myptr;
by id id_2;
* When this is the first observation in a BY-group, save the
* current observation number (pointer).
* Also initialize a flag variable that will become 1 if any
* obs contains a VALUE of 0;
* The variables are in a RETAIN statement, so they keep their
* values as the SET statement above is executed for each obs
* in the BY-group.;
if first.id_2
then do;
startptr=myptr;
haszero=0;
end;
* This statement is executed for each observation. We check
* whether VALUE is 0 and, if so, record that fact.;
if value = 0
then haszero=1;
* At the end of the BY-group, we check to see if there were
* any observations with VALUE = 0. If so, we go back using
* another SET statement, re-read them via direct access, and
* write them to the output dataset.
* (Note that if VALUE order is not relevant, you can gain a bit
* more efficiency by writing the current obs first, then going
* back to get the rest.);
if last.id_2 and haszero
then do;
* When LAST and FIRST at the same time, there is only one
* obs, so no need to backtrack, just output and go on.;
if first.id_2
then output obstokeep;
else do;
* Here we assume that the observations are sequential
* (which they will be for a sequential SET statement),
* so we re-read these observations using another SET
* statement with the POINT option for direct access
* starting with the first obs of the by-group (the
* saved pointer) and ending with the current one (the
* current pointer).;
do i=startptr to myptr;
set x point=i;
output obstokeep;
end;
end;
end;
run;
proc sql;
select a.*,b.value from (select id,id_2 from have where value=0)a left join have b
on a.id=b.id and a.id_2=b.id_2;
quit;
I currently have a health injury data set of scores 0-6, where 0 is no injury and 6 is fatal injury. This is across 6 categorical body region variables. I'm attempting to construct an Abbreviated Injury Scale, where the three highest scores in an observation would be considered for the calculations. How do I filter the three highest in each row in SAS? Below is an example:
ID A B C D E F
1 0 0 0 3 4 0
2 1 2 1 4 0 0
3 0 0 5 0 0 0
4 1 2 1 5 4 0
So in OBS 1, scores 3, 4, and 0 would be used; OBS 2 - 4, 2, and 1; OBS 3 - 5, 0, and 0; OBS 4 - 5, 4, 2.
I've provided code below to do what you asked, and detailed out the steps enough that you should be able to modify it for many options/uses.
Basically, it takes your data, transposes it as Quentin suggested and then uses proc means to output the top 3 observations for each ID.
DATA NEW;
INPUT ID A B C D E F;
CARDS;
1 0 0 0 3 4 0
2 1 2 1 4 0 0
3 0 0 5 0 0 0
4 1 2 1 5 4 0
RUN;
PROC TRANSPOSE DATA=NEW OUT=T_OUT(RENAME=(_NAME_ = VARIABLE COL1=VALUES));
BY ID;
VAR A B C D E F;
PROC PRINT DATA=T_OUT;
RUN;
PROC MEANS DATA=T_OUT NOPRINT;
CLASS ID;
TYPES ID;
VAR VALUES;
OUTPUT OUT=TOP3LIST(RENAME=(_FREQ_=RANK VALUES_MEAN=INDEX_CRITERIA))SUM= MEAN=
IDGROUP(MAX(VALUES) OUT[3] (VALUES VARIABLE)=)/AUTOLABEL AUTONAME;
PROC PRINT DATA=TOP3LIST;
RUN;
***THEN YOU CAN MERGE THIS DATA SET TO YOUR ORIGINAL ONE BY ID TO GET YOUR INDEX CRITERIA ADDED TO IT***;
***THE INDEX_CRITERIA IS A MEAN FROM PROC MEANS BEFORE THE KEEPING OF JUST THE TOP3 VALUES***;
DATA FINAL (DROP=_TYPE_ RANK VALUES_Sum VALUES_1 VALUES_2 VALUES_3 VARIABLE_1 VARIABLE_2 VARIABLE_3);
MERGE NEW TOP3LIST;
INDEX_CRITERIA2=SUM(VALUES_1, VALUES_2, VALUES_3)/3; *THIS CRITERIA IS AVERAGE OF THE KEPT 3 VALUES;
BY ID;
PROC PRINT DATA=FINAL;
RUN;
Best regards,
john