Error in calculating std of the residuals in each regression in each simulation

oriti — Sun, 27 Oct 2013 13:11:26 GMT

Dear all,

I enclosed a program which simulate regressions

that you can try to run.

I succeeded in calculating the coefficient matrix of the regression (b).

but when I try to calculate the residuals I get just one column in the matrix that is correct but the other columns are with ones (1).

so the std of the residuals is also is not correct. (bolded with blue colour in the program)

I don't understand why that happens..

Who can help?

Many thanks...

proc iml;

n=10;

b=j(6,n);

**do j=1 to n; *Number of simulations;**

resid=j(239,n);

stdresid=j(n,1);

xtl = j(240, 1);

xsl=j(240, 1);

y=j(240, 1);

xtl[1,1]=0.926145577;

xsl[1,1] = 1.367164758;

y[1,1] = -27.99208409;

c=0.24269;

e=0.02106;

d=0.97564;

h=-0.24776;

m=-0.00362;

l=-0.00111;

q=0.00091339;

r=1.00339;

/ specify the mean and covariance of the population /

Mean = {0, 0, 0};

Cov = {0.0343693345 0.0005718713 -.0000330633,

0.0005718713 0.0008609739 0.0000083048,

-.0000330633 0.0000083048 0.0000022582};

NumSamples = 240;

call randseed(0); / set seed for the RandNormal module /

U = RandNormal(NumSamples, Mean, Cov);

do i=2 to 240;

Y[i,1]=y[1,1]+u[i,1];

**xsl[i,1]=c+ey[(i-1),1]+dxsl[(i-1),1]+h*xtl[(i-1),1]+u[i,2];**

**xtl[i,1]=m+ly[(i-1),1]+qxsl[(i-1),1]+r*xtl[(i-1),1]+u[i,3];**

end;

ylag=lag(y);

xsl2=xsl##2;

xtl2=xtl##2;

**/* Step 1: Compute X`X and X`y */**

**x = j(nrow(y), 1, 1) || ylag || xsl|| xtl || xsl2||xtl2; /* add intercept column */**

**v = {1}; /* specify rows to exclude */**

**idx = setdif(1:nrow(x), v); /* exclude values in v */**

**xnew = x[idx, ]; /* extract submatrix */**

**xpx = xnew`* xnew; /* cross-products */**

**v = {1}; /* specify rows to exclude */**

**idy = setdif(1:nrow(y), v); /exclude values in v /**

**ynew = y[idx, ]; /* extract submatrix*/**

**xpy = xnew` * ynew;**

b[,j]= solve(xpx, xpy);

**resid[,j]=ynew-xnew*b[,j];**

stdresid[j,]=std(resid[,j]);

end;

print b;

print resid;

print stdresid;

quit;

Re: Error in calculating std of the residuals in each regression in each simulation

Rick_SAS — Mon, 28 Oct 2013 19:48:16 GMT

Move the initialization of RESID and STDRESID outside of the DO loop for J.

Re: Error in calculating std of the residuals in each regression in each simulation

oriti — Tue, 29 Oct 2013 06:26:34 GMT

Rick, thank you!

topic Re: Error in calculating std of the residuals in each regression in each simulation in SAS/IML Software and Matrix Computations

Error in calculating std of the residuals in each regression in each simulation

proc iml;

n=10;

b=j(6,n);

do j=1 to n; *Number of simulations;

resid=j(239,n);

stdresid=j(n,1);

xtl = j(240, 1);

xsl=j(240, 1);

y=j(240, 1);

xtl[1,1]=0.926145577;

xsl[1,1] = 1.367164758;

y[1,1] = -27.99208409;

c=0.24269;

e=0.02106;

d=0.97564;

h=-0.24776;

m=-0.00362;

l=-0.00111;

q=0.00091339;

r=1.00339;

/** specify the mean and covariance of the population **/

Mean = {0, 0, 0};

Cov = {0.0343693345 0.0005718713 -.0000330633,

0.0005718713 0.0008609739 0.0000083048,

-.0000330633 0.0000083048 0.0000022582};

NumSamples = 240;

call randseed(0); /** set seed for the RandNormal module **/

U = RandNormal(NumSamples, Mean, Cov);

do i=2 to 240;

Y[i,1]=y[1,1]+u[i,1];

xsl[i,1]=c+e*y[(i-1),1]+d*xsl[(i-1),1]+h*xtl[(i-1),1]+u[i,2];

xtl[i,1]=m+l*y[(i-1),1]+q*xsl[(i-1),1]+r*xtl[(i-1),1]+u[i,3];

end;

ylag=lag(y);

xsl2=xsl##2;

xtl2=xtl##2;

/* Step 1: Compute X`X and X`y */

x = j(nrow(y), 1, 1) || ylag || xsl|| xtl || xsl2||xtl2; /* add intercept column */

v = {1}; /* specify rows to exclude */

idx = setdif(1:nrow(x), v); /* exclude values in v */

xnew = x[idx, ]; /* extract submatrix */

xpx = xnew`* xnew; /* cross-products */

v = {1}; /* specify rows to exclude */

idy = setdif(1:nrow(y), v); /*exclude values in v */

ynew = y[idx, ]; /* extract submatrix*/

xpy = xnew` * ynew;

b[,j]= solve(xpx, xpy);

resid[,j]=ynew-xnew*b[,j];

stdresid[j,]=std(resid[,j]);

end;

print b;

print resid;

print stdresid;

quit;

Re: Error in calculating std of the residuals in each regression in each simulation

Re: Error in calculating std of the residuals in each regression in each simulation

**do j=1 to n; *Number of simulations;**

/ specify the mean and covariance of the population /

call randseed(0); / set seed for the RandNormal module /

**xsl[i,1]=c+ey[(i-1),1]+dxsl[(i-1),1]+h*xtl[(i-1),1]+u[i,2];**

**xtl[i,1]=m+ly[(i-1),1]+qxsl[(i-1),1]+r*xtl[(i-1),1]+u[i,3];**

**/* Step 1: Compute X`X and X`y */**

**x = j(nrow(y), 1, 1) || ylag || xsl|| xtl || xsl2||xtl2; /* add intercept column */**

**v = {1}; /* specify rows to exclude */**

**idx = setdif(1:nrow(x), v); /* exclude values in v */**

**xnew = x[idx, ]; /* extract submatrix */**

**xpx = xnew`* xnew; /* cross-products */**

**v = {1}; /* specify rows to exclude */**

**idy = setdif(1:nrow(y), v); /exclude values in v /**

**ynew = y[idx, ]; /* extract submatrix*/**

**xpy = xnew` * ynew;**

**resid[,j]=ynew-xnew*b[,j];**