It works for this:   con pvalueGE{i in pool, j in groups: pvalue[i]<0.1}: x[i,j]=0;  Thank you! ... View more

I will try it later today to see if I can work it out. SAS/OR is difficult for beginner.   Thanks for the example. What is the second approach in details?   ... View more

To say if I want to create a quantitative constraint "choose items from the dataset with pvalue>=0.1" , how should I express this constraint in SAS/OR? This time the question should be clear. Thank you! All the constraints in this example used "sum", tried a few times but failed.  ... View more

Rob, I noticed for this example (in 2-dimension), almost every constraint used "SUM", but if I want to define every single value in a linear program, e.g., pvalue[i]<0.2, how should I create this constraint?  Thank you！ ... View more

I have chosen one of your previous solutions (2 dimension's)  as the best solution.Thank you so much!   My initial interest for SAS/OR is detecting IIS, remember you replied my first post here: https://communities.sas.com/t5/Mathematical-Optimization/SAS-OR-in-LP-MIP-which-method-is-used-for-detecting-IIS/m-p/495128#M2394 I am a quantitative researcher but have limit knowledge in OR(this is the first time project), So I didn't really know IIS is for LP (not MILP). Thank you for telling this. I have several addition categorical constraints to add to the model, which I know will cause infeasibility. In addition to your code MILP version, I figured out for LP: proc optmodel presolver=none; .................... the code your developed; .................... solve with lp relaxint;/ iis=on; It seems to stop relaxing constraints;    There is another problem, how many nodes are normal? when I add constraint SD, for group 1..3, it worked fine; for group=4, it slowed down a little bit; when group=5, it ran with thousands of nodes and could not give a solution (maybe there wasn't one). I had to stop the program manually. Is there anyway to avoid this?    ... View more

Hello RobPratt. I have tested your code, it worked greatl. Thank you so much. For my question 2 using SAS macro, I tried using %let group=5, group 1..&group.; it worked. For my question 1: I still have some details to work on, should I consider the objective function you wrote, means to minimize each y-value in group 1..5?             min z = sum{j in groups} y[j];<code></code>     I tried both using and not using "decomp" option, it made a difference in speed. For the last statement using "decomp", can I also have “iis=on”？Are they comparable? I wish to use MILP/branch and bound/indicate IIS for my project (is that possible?)          solve with milp /decomp iis=on;<code></code>          I tried to replace “decomp” with "IIS=on"(line 384), it gave me the following error:  384 solve with milp /decomp  iis=on; --- 688 22 ERROR 688-782: The option IIS is unrecognized. ERROR 22-322: Expecting one of the following: ABSOBJGAP, ALLCUTS, CONFLICTSEARCH, CUTCLIQUE, CUTFLOWCOVER, CUTFLOWPATH, CUTGOMORY, CUTGUB, CUTIMPLIED, CUTKNAPSACK, CUTLAP, CUTMILIFTED, CUTMIR, CUTOFF, CUTS, CUTSFACTOR, CUTSTRATEGY, CUTZEROHALF, DECOMP, DECOMP_MASTER, DECOMP_MASTER_IP, DECOMP_SUBPROB, EMPHASIS, FEASTOL, HEURISTICS, INTTOL, LOGFREQ, LOGLEVEL, MAXNODES, MAXSOLS, MAXTIME, NODESEL, NOPRIMALIN, OPTTOL, PRESOLVER, PRIMALIN, PRINTFREQ, PRINTLEVEL2, PRIORITY, PROBE, RELOBJGAP, RESTARTS, SCALE, SEED, STRONGITER, STRONGLEN, SYMMETRY, TARGET, TIMETYPE, VARSEL.   when I switched to  LP, it relax all the constraints and did not pick any x[i] for me. solve with lp / iis=on;   WARNING: The problem contains integer variables. The relaxed LP will be solved. Specify the RELAXINT keyword on the SOLVE statement to avoid this warning. NOTE: The IIS= option is enabled. Objective Phase Iteration Value Time P 1 1 1.774000E+03 0 P 1 138 1.121796E+02 0 P 1 233 8.137021E+00 0 P 1 328 3.910344E-01 0 P 1 358 0.000000E+00 0 P 1 359 0.000000E+00 0 NOTE: The IIS= option found this problem to be feasible. NOTE: The IIS solve time is 0.02 seconds. ... View more

Thank you so much @RobPratt I will continue developing the code for my project. I will update how it goes.  ... View more

Thank you soooooo much!!!   After last post, I have figured out the mistakes I made, and corrected the computation of SD term. But what you wrote is much better than mine.   Here I have 2 extra questions: Question 1. For a further modification of this program, e.g. to add a dimension, the output will be like this: [1]       x1   x2  x3  ....  x5; 00001  0     0   1  ....    0 00002  0    1   0   ....    0 00003  1    0   0   ....    0   .................. .................. 00450  0  1   0   ......    0; which means I wish to select multiple groups x[i,j ] require: for each group( in 1 to j) , choose 50 item from the pool:sum(x[ i ])=50; each item can only goes into one group: sum(x)<=1, no overlap;   I created another dataset named "group" with only 1 variable (x1-x5), but I could not read x1-x5 into the dataset [pool] to make another dimension, I also don't know how to modify the program in "proc optmodel";   Question 2: can "proc optmodel" be used in SAS Macro? e.g. %macro forms (group= ); data xxx; set xxx; .........; run;   proc optmodel; .......; .......; .......;   %mend; %forms (group=5);    Thank you!!! Appreciate!!! ... View more