I need your remarks in this problem.
I have ode. diff(y(x),x)=f(x,y); x in [0,a]; h:=a/(2*N); stepsize.
When the the true solution is not Known, we can test the rate of convergence, of numerical solution. The Numerical solution generated when the stepsize is 2*h denoted by y_i^(2h) and the numerical solution with step size h will be denoted by y_i^(h).
if we define the epsilon(h):=sqrt ( 1/(N+1)*add( (y_i^(2h) -y_(2i)^(h) )^2, i=0..N));
If we useForward Euler ( it's Known that the golbal error isof order 1 and local error of order 2) in the case when the exact solution is know.
But, If we use epsilon(h), and for the same method can some one know the order of Error =h^?????.