Thanks @Ksharp , now the error is  ERROR: (execution) Invalid subscript or subscript out of range. Additionally, how can change the IML code to caculate z value for each observation (more like the IML equivalent of "diff" command in OPTMODEL?   ... View more

Hello @RobPratt , @Ksharp    I have written the following code but am getting error that  ERROR: (execution) Matrix has not been set to a value   My code is given below:   proc iml; /*start a function to minimized*/ start obj_z(y) ; num_col=ncol(y); num_row=nrow(y); num_rat=y[,7]; do i = 1 to num_row; do j= 1 to (num_col-2); p[i,j] = CDF('Normal', y[i,j+1]) - CDF('NORMAL', y[i,j]); phi[i,j] = CDF('Normal', (y[i,j]-sqrt(rho)*Z)/sqrt(1-rho)); Delta [i,j] = Ph([i,j+1] - Phi[i,j]); sum_z= sum((num_r (i)* (P[i,j] - Delta[i,j])**2) / (Delta[i,j] * (1 - Delta[i,j]))); return (sum_z); end; end; finish obj_z ; x_1={0 -1.78 -2.68 -3.06 -3.4 -3.83 1885, 0 2.12 -2.15 -2.63 -2.94 -3.34 3151, 0 2.45 1.16 -1.39 -1.99 -2.54 246, 0 2.48 1.92 1.19 -1.13 -2.39 80, 0 2.68 2.32 1.82 1.52 -1.56 89} ; num_row_1 = nrow(x_1); num_col_1 = ncol(x_1); /*print n_x p_x ;*/ varNames= ("cat3" : "cat8" || "num_rat"); rowNames= ("cat3" : "cat7"); create y from x_1[colname=varNames rowname=rowNames]; append from x_1 [rowname = rowNames]; /*close x_1;*/ start_z= {0.89}; /*initialize z*/ rho={0.8}; /* rho is a constant*/ OPTN= {0 2}; /* options for minimimization problem and result output*/ call NLPTR(xres, rc, "obj_z",start_z , OPTN) ; /* calling the function*/ /*print "result";*/ Quit;   ... View more

Hello everyone,   Thank you for your code especially Rob. I do not have proc optmodel in my SAS subscription, but I have SAS IML. I have tried to write the code in IML but failed miserably. I tried to read from SAS documentation and online topics and have written the following code. I would be very thankful if you guys can help me fix it.   proc iml; x={0 -1.78 -2.68 -3.06 -3.4 -3.83 1885, 0 2.12 -2.15 -2.63 -2.94 -3.34 3151, 0 2.45 1.16 -1.39 -1.99 -2.54 246, 0 2.48 1.92 1.19 -1.13 -2.39 80, 0 2.68 2.32 1.82 1.52 -1.56 89} ; varNames= ("cat3" : "cat8" ||"num_rat"); rowNames= ("cat_3" : "cat_7"); n_x = nrow(x); p_x = ncol(x-1); print n_x p_x ; create mydata1 from x[colname=varNames rowname=rowNames]; append from x [rowname = rowNames]; close mydata1; start obj_z(p, phi, delta, rho, init_z); num_col=ncol(x-1); num_row=nrow(x); num_r=x[,7]; do i = 1 to num_row; p{i in num_rows, j in num_col} = CDF('Normal', x[i,j+1]) - CDF('NORMAL', x[i,j]); phi{i in num_rows, j in num_col} = CDF('Normal', (Cat_C[i,j]-sqrt(&rho)*Z)/sqrt(1-&rho)); delta {i in num_rows, j in num_col} = Phi[i,j+1] - Phi[i,j]; sum_z= (num_rat[i] * (P[i,j] - Delta[i,j])**2 / (Delta[i,j] * (1 - Delta[i,j]))); finish; call nlpqn(min(obj_z)) = "result"; print "result"; run; ... View more

Thanks Ksharp. My problem is that if most of my group friends also don't have Proc Optmodel, thats why I would be very thankful if you guys can help me with the code in SAS IML.   Regards, Jass  ... View more

I have IML.  ... View more

Thanks Rob, you are a genius. your solution works nicely.  I have one problem, if I don't have proc optmodel in my SAS Enterprise guide, what other options can be used to perform the same function?   Thanks & Regards, Jass ... View more

https://www.z-riskengine.com/media/1032/a-one-parameter-representation-of-credit-risk-and-transition-matrices.pdf ... View more

I understand Brad, and apologize for not being clear earlier. Below is the equation I need Minimize:   where   and  The matrix is data test ; data test ; input cat $ cat3 cat4 cat5 cat6 cat7 cat8 ; cards; cat3 0 -1.78 -2.68 -3.06 -3.4 -3.83 cat4 0 2.12 -2.15 -2.63 -2.94 -3.34 cat5 0 2.45 1.16 -1.39 -1.99 -2.54 cat6 0 2.48 1.92 1.19 -1.13 -2.39 cat7 0 2.68 2.32 1.82 1.52 -1.56 ; run; %let rho = 0.8; *It will be changed after every run*; %let z= 0.89; * It should be minimized*; DATA TEST1; SET test; Array Cat_C(6) cat3-cat8; Array Cat_g(5) catg3-catg7; do i= 1to 5; cat_g(i)= num_rat * ((((CDF('Normal', Cat_C(i+1))-CDF('NORMAL',Cat_C(i))- ((CDF('NORMAL', ((Cat_C(i+1) - (sqrt(&rho)*&z))/sqrt(1-&rho))))-(CDF ('NORMAL', ((Cat_C(i) - (sqrt(&rho)*&z))/sqrt(1-&rho))))))**2))/ (((CDF('NORMAL', ((Cat_C(i+1) - (sqrt(&rho)*&z))/sqrt(1-&rho))))-(CDF ('NORMAL', ((Cat_C(i) - (sqrt(&rho)*&z))/sqrt(1-&rho)))))* (1-((CDF('NORMAL', ((Cat_C(i+1) - (sqrt(&rho)*&z))/sqrt(1-&rho))))-(CDF ('NORMAL', ((Cat_C(i) - (sqrt(&rho)*&z))/sqrt(1-&rho)))))))); end; run;   where x(g+1) is the movement from cat3 to cat4. and rho needs to be any fixed number (say 0.8) and z needs to be minimized using any initial number (say 0.89).  I hope it helps. Thanks again. ... View more

Thanks Paige, but I think this is not the required answer. I expect the do loop to run through each variable (columns) and calculate the function. Also I intend to minimize the function to one variable (lets say a z_score which can be said to (0 + -1.78)*1885*Z + (-1.78 + -2.68)*1885 *Z + (-2.68 + -3.06)*1885*Z + (-3.06 + -3.4)*1885*Z + (-3.4 + -3.83)*1885*Z. Sorry I should have mentioned "Z" in my original post. ... View more