Reverse function of group by

[skallamp]

Hi,

I have a particular scenario where I wanted to get the reverse of group by function.

Here is the sample data

Table 1

medication	ndc	rx_count	total_qty
Lipitor	59148000813	1	10
Lipitor	59148000813	1	10
Lipitor	59148000813	1	10
Lipitor	59148000813	1	10
Lipitor	59148000813	1	10
Lipitor	59148000813	1	10
Lipitor	59148000813	1	10
Lipitor	59148000813	1	10
Lipitor	59148000813	1	10
Lipitor	59148000813	1	10
Lipitor	59148000813	-1	-10
Lipitor	59148000813	1	10

If I do the summary using the below code

PROC SQL;

CREATE TABLE WORK.QUERY_FOR_TESTTEST AS

SELECT t1.medication,

t1.ndc,

(SUM(t1.rx_count)) FORMAT=BEST12. AS SUM_of_rx_count,

(SUM(t1.total_qty)) FORMAT=BEST12. AS SUM_of_total_qty

FROM WORK.TESTTEST AS t1

GROUP BY t1.medication, t1.ndc;

QUIT;

I get the result as below.

Table 2

medication	ndc	SUM_of_rx_count	SUM_of_total_qty
Lipitor	59148000813	10	100

Is there a way that I get the Table 1 as out put from table 2. Table 2 is the aggregated data, what I want is the extended as table 1.

What would be the query I should use?

Thanks,

+Sathyan

[Ksharp]

You want add this aggregated data at the bottom of Table1 ?

PROC SQL;

CREATE TABLE WORK.QUERY_FOR_TESTTEST AS select * from WORK.TESTTEST union all corr

SELECT t1.medication,

t1.ndc,

(SUM(t1.rx_count)) FORMAT=BEST12. AS SUM_of_rx_count,

(SUM(t1.total_qty)) FORMAT=BEST12. AS SUM_of_total_qty

FROM WORK.TESTTEST AS t1

GROUP BY t1.medication, t1.ndc;

QUIT;

[skallamp]

I do not want to add the result at the bottom.

What I want is, how do we derive the table 1 detailed information from table 2 summary? I am OK if we do not get the negative values.

[LinusH]

Just join the two tables using the group by columns as join key.

[andreas_lds]

skallamp wrote:

javascript:void(0);
I do not want to add the result at the bottom.
What I want is, how do we derive the table 1 detailed information from table 2 summary? I am OK if we do not get the negative values.

Possible only if "total_qty" is constant for all obs in one group and no negative values are present. With a least one negative total_qty/rx_count the number of "restored" observations won't match the number of obs in the original dataset.

data work.faked_original;

       set work.QUERY_FOR_TESTTEST;

       length rx_count total_qty i 8;

       drop sum_of_: i;

       do i = 1 to sum_of_rx_count;

              rx_count = 1;

              total_qty = sum_of_total_qty / sum_of_rx_count;

              output;

       end;

run;

[skallamp]

Medication	NDC	RX Count	Total QTY
Abilify 10mg Tablet	24236027902	1	30
Abilify 10mg Tablet	59148000813	257	2297
Abilify 15mg Tablet	59148000913	19	217.5
Abilify 20mg Tablet	59148001013	441	8145.5

Hi Andreas,

Thanks, that was helpful.

But that code will fail when the case is as above. Rx count need not be 1 always.

Let me explain. In line 2 (10mg Tablet) i need 257 identical lines as output, however Total qty in output should be 8.93 (i.e 2297 / 257)

How do we crack this?

Regards,

+Sathyan

[Astounding]

You will have to set up a few more rules about how to do it.  Once you allow that negative numbers could have been part of the unsummarized table, there are an infinite number of solutions to the problem.

[joehinson]

Is that really what you want?

data have;

input @1 Medication $ @21 NDC $ @33 sRX_Count @37 sTotal_QTY;

datalines;

Abilify 10mg Tablet 24236027902 1   30

Abilify 10mg Tablet 59148000813 257 2297

Abilify 15mg Tablet 59148000913 19  217.5

Abilify 20mg Tablet 59148001013 441 8145.5

;

run;

data want(keep=Medication NDC RX_Count Total_QTY) ;

set have;

itx=int(sTotal_QTY/sRX_Count);

do count=1 to itx;

Total_QTY=sTotal_QTY/sRX_Count;

RX_Count=sRX_Count/itx;

output;

end;

run