Hello,
I wrote the following code that determines the weights maximizing a ratio. These weights should be positive given the constraint on variables:
var W{Assets_BY} >= 0 <= 0.1;
but many of them are actually negative (admittedly small, with order E-7). However, when I rescale them so that their sum is equal to 1, they are no longer negligible with the smaller value being equal to -0.135.
%let byvar = grp;
proc optmodel printlevel=0;
set OBS;
num grp {OBS};
set <string> Assets;
set <num,str> GROUPS_ASSETS;
number MVC{GROUPS_ASSETS, Assets};
read data Work.MVC into Assets=[Name];
read data Work.MVC into OBS=[_N_] grp;
read data Work.MVC nomiss into GROUPS_ASSETS=[k=grp i=Name] {j in Assets} <MVC[k,i,j]=col(j)>;
set BYSET = setof {i in OBS} &byvar.[i];
num by;
set OBS_BY = {i in OBS: &byvar.[i] = by};
set Assets_BY = setof {o in OBS_BY, <(grp[o]),a> in GROUPS_ASSETS} a;
var W{Assets_BY} >= 0 <= 0.1;
impvar Denom = sqrt(sum{i in Assets_BY, j in Assets_BY}W[i]*MVC[by,i,j]*W[j]);
maximize DR = log(sum{i in Assets_BY}W[i]*sqrt(MVC[by,i,i]) / Denom);
num W_sol {GROUPS_ASSETS};
do by = BYSET;
put by=;
solve;
for {i in Assets_BY} W_sol[by,i] = W[i].sol;
end;
create data Work.Weights_MD from [&byvar i] W=W_sol;
quit;
Thank you for any advice to obtain weights that respect the lower and upper bounds,
Best,
PS: if necessary, I will provide a small working example
The default feasibility tolerance is 1e-6, so 1e-7 is within that tolerance. You can change the value by using the FEASTOL= option.
If you want the sum of W to be 1, it might be best to explicitly declare such a constraint. If that doesn't suffice, please share your data.
The default feasibility tolerance is 1e-6, so 1e-7 is within that tolerance. You can change the value by using the FEASTOL= option.
If you want the sum of W to be 1, it might be best to explicitly declare such a constraint. If that doesn't suffice, please share your data.
Thank you very much for your help 🙂
I will try to define a lower feasibility tolerance. As for the "sum of weight" constraint, I removed it because it was much longer with it, but perhaps it will be more efficient than a lower feasibility tolerance. I will perform some tests. Thank you!
Registration is open! SAS is returning to Vegas for an AI and analytics experience like no other! Whether you're an executive, manager, end user or SAS partner, SAS Innovate is designed for everyone on your team. Register for just $495 by 12/31/2023.
If you are interested in speaking, there is still time to submit a session idea. More details are posted on the website.
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.