how to do regression in rows and interpolation in columns in 2d matrices

i am using matlab to do interpolation and regression in 3 different matrices with the same size. the matrices call angle, rotation and error.

i need to do the 3rd order polynomial regression in rows of rotation(=x) and error(=y) matrices and save the new error data az y', then do the linear interpolation in column of angle(=x) and y' matrices and at the end plot the (angle(=x), rotation(=y), error(=z)) in 3d space.

for the regression i am using polyfit and polyval functions and find the polynomial coefficent matrix and y' matrix but for the second step when i use F=TriScatteredInterp(x,y,z) this function interpolate the data in 2 dimention, it means the points that i got from regression are changing , but i dont want that.i wanna the regression that i have done in the rows and acquired points dont change.

could you please help me with that, how can i do the regression in rows and interpolation in columns and plot the graph in much more detail than the data points that i have?

here is the last part of my program, where i have used the funcion F:

x=transpose(angle);
y=transpose(rotation);
z=transpose(Metx1);   % error function after regression

F = TriScatteredInterp(x,z);

ti = -150:1:150;
ti1 = -40:1:40;
[qx,qy] = meshgrid(ti1,ti);
qz = F(qx,qy)
surf(qx,qy,qz);
colormap bone
hold on;
stem3(x,y,z,'fill','r');