BookmarkSubscribeRSS Feed
🔒 This topic is solved and locked. Need further help from the community? Please sign in and ask a new question.
Decay2020
Fluorite | Level 6

Hi all,

I have 3 inputs, one data "Have" and two values in macro variable "value1" and "value2". I am interested in getting the data "want" as per the condition given below. Thanks in advance.

 

data have;
infile cards delimiter=",";
input col1 col2 col3;
cards;
1, 0.01, 100

2, 0.02, 90

3, 0.03, 80

4, 0.04, 70

5, 0.02, 60

6, 0.06, 50

7, 0.07, 40

;

run;

%let value1 = -4.608753929;

%let value2 = 0.28179071;

now, we want to calculate col4 as value1 + (value2*col1)

and col5 as EXP(col4)/(1+EXP(col4)) such that

sumproduct(col5, col3)/sum(col3) = sumproduct(col2,col3)/sum(col3) by changing the value1 and keeping the value2 same.

 

After optimization the "value1" should be -4.56266639 (computed in excel) and col4, col5 should be as given in the table below:

 

Data Want

col1col2col3col4col5
1  0.01100-4.280875680.013641871
2  0.0290-3.999084970.018002379
3  0.0380-3.717294260.023723163
4  0.0470-3.435503550.031204128
5  0.0260-3.153712840.040945231
6  0.0650-2.871922130.053559135
7  0.0740-2.590131420.069776252
1 ACCEPTED SOLUTION

Accepted Solutions
RobPratt
SAS Super FREQ

The following does what you want:

proc optmodel;
   set OBS;
   num col1 {OBS};
   num col2 {OBS};
   num col3 {OBS};
   read data have into OBS=[_N_] col1 col2 col3;

   var value1 init &value1;
   num value2 = &value2;
   impvar col4 {i in OBS} = value1 + value2 * col1[i];
   impvar col5 {i in OBS} = EXP(col4[i]) / (1 + EXP(col4[i]));
   con SumProduct:
      sum {i in OBS} col5[i] * col3[i] = sum {i in OBS} col2[i] * col3[i];

   solve noobj;
   print value1;
   print col1 col2 col3 col4 col5;
   create data want(drop=i) from [i] col1 col2 col3 col4 col5;
quit;
value1
-4.5621

[1] col1 col2 col3 col4 col5
1 1 0.01 100 -4.2803 0.013649
2 2 0.02 90 -3.9985 0.018012
3 3 0.03 80 -3.7168 0.023736
4 4 0.04 70 -3.4350 0.031220
5 5 0.02 60 -3.1532 0.040966
6 6 0.06 50 -2.8714 0.053586
7 7 0.07 40 -2.5896 0.069811

View solution in original post

8 REPLIES 8
blueskyxyz
Lapis Lazuli | Level 10
%let value1 = -4.608753929;
%let value2 = 0.28179071;

data want;
	retain sumproduct1 sumproduct2 sum_col3 0;
	set have;
	col4 =&value1. + (&value2.*col1);
	col5=EXP(col4)/(1+EXP(col4));
    sumproduct1=sumproduct1+col5*col3;  col5_col3=col5*col3;
    sumproduct2=sumproduct2+col2*col3;  col2_col3=col2*col3;
	sum_col3=sum_col3+col3;

	sumproduct1_=sumproduct1/sum_col3; 
	sumproduct2_=sumproduct2/sum_col3; 
	xx=col4-col5;
run;

sumproduct(col5, col3)/sum(col3) = sumproduct(col2,col3)/sum(col3)??

Decay2020
Fluorite | Level 6
i need to change the value1 from -4.608753929 such that sum product(col5 and col3)/ sum(col3) = 0.030612

0.030612 is sumproduct(col2,col3)/sum(col3) which is computed in excel.
blueskyxyz
Lapis Lazuli | Level 10
proc sql;
    create table sumproduct as
    select -4.608753929 as value1
			,0.28179071 as value2
			,col1,col2,col3
			, calculated value1 + (calculated value2 *col1) as col4
			, EXP( calculated col4)/(1+EXP(calculated col4))as col5
			, sum(calculated col5*col3)/sum(col3)  as sum1, sum(col2*col3)/sum(col3)  as sum2 
    from have;
quit;

/*need a do-loop for value1?*/ data sump; set sumproduct; if sum1=sum2 then value1=-4.56266639; run;
Decay2020
Fluorite | Level 6
No, we don't have value1 = -4.56266639.
we need to optimize the value1 starting from -4.608753929 such that sumproduct(col5, col3)/sum(col3) = 0.030612
RobPratt
SAS Super FREQ

The following does what you want:

proc optmodel;
   set OBS;
   num col1 {OBS};
   num col2 {OBS};
   num col3 {OBS};
   read data have into OBS=[_N_] col1 col2 col3;

   var value1 init &value1;
   num value2 = &value2;
   impvar col4 {i in OBS} = value1 + value2 * col1[i];
   impvar col5 {i in OBS} = EXP(col4[i]) / (1 + EXP(col4[i]));
   con SumProduct:
      sum {i in OBS} col5[i] * col3[i] = sum {i in OBS} col2[i] * col3[i];

   solve noobj;
   print value1;
   print col1 col2 col3 col4 col5;
   create data want(drop=i) from [i] col1 col2 col3 col4 col5;
quit;
value1
-4.5621

[1] col1 col2 col3 col4 col5
1 1 0.01 100 -4.2803 0.013649
2 2 0.02 90 -3.9985 0.018012
3 3 0.03 80 -3.7168 0.023736
4 4 0.04 70 -3.4350 0.031220
5 5 0.02 60 -3.1532 0.040966
6 6 0.06 50 -2.8714 0.053586
7 7 0.07 40 -2.5896 0.069811
Decay2020
Fluorite | Level 6
Thank you @RobPratt.. this solution works perfectly. But can we update the constraint:
sum {i in OBS} col5[i] * col3[i] = sum {i in OBS} col2[i] * col3[i];
into :
sumproduct(col5, col3)/sum(col3) = sumproduct(col2,col3)/sum(col3) somehow?
as we need to divide the sumproduct with sum of col3.
I understand it wont impact the result as it is cancelling out, but still needed.
RobPratt
SAS Super FREQ
   con SumProduct:
      (sum {i in OBS} col5[i] * col3[i]) / (sum {i in OBS} col3[i]) 
    = (sum {i in OBS} col2[i] * col3[i]) / (sum {i in OBS} col3[i]);
blueskyxyz
Lapis Lazuli | Level 10
proc optmodel, a new proc step for me, thanks @RobPratt, sas/or never be used in the pharmaceutical industry

sas-innovate-2024.png

Available on demand!

Missed SAS Innovate Las Vegas? Watch all the action for free! View the keynotes, general sessions and 22 breakouts on demand.

 

Register now!

Multiple Linear Regression in SAS

Learn how to run multiple linear regression models with and without interactions, presented by SAS user Alex Chaplin.

Find more tutorials on the SAS Users YouTube channel.

Discussion stats
  • 8 replies
  • 902 views
  • 2 likes
  • 3 in conversation