here's some untested pseudo code that might fit what you are describing
for( par1 = 1, par1 <= 3, par1+=1, // 3...how ever many combinations there are
for(par2 = 1, par2 <= 5, par2 +=1, // 5
for(par3 = 1, par2 <= 2, par3+= 1, // 2
/* get the vectors from matlab using par1,par2,par3 */
/* somehow mat_t, mat_y1, and mat_y2 are now JSL matrices, n rows by 1 col */
for( i = 1, i<= n, i+=1, // n is the vector length, write a row for each element
dt<<addrows(1);
dt:y1 = mat_y1[i]; // different values for n rows. also for y2
dt:t = mat_t[i];
dt:p1 = par1; // repeats same value for n rows. also for p2, p3
... close all the parens ...
I named the table variable p1 for par1 to simplify the scoping issues when column names are the same as variable names. I imagined you make separate calls to matlab for each combination of par1,2,3. I imagined those vectors show up as JSL matrices. Factors and Designs and matlab are not my strong area, I may be misunderstanding the issue.
Craige