a) . If $ϕ(h) = L − c_1h^{1/2} − c_2h^{2/2} − c_3h^{3/2} − ···$, then what combination of $ϕ(h)$ and $ϕ(h/2)$ should give an accurate estimate of L?
b) If $ϕ(h) = L − c_1h − c_2h^2 − c_3h^3 − ···$, then what combination of $ϕ(h)$ and $ϕ(h/2)$ should give an accurate estimate of L?
The solution given in the textbook for b is $L=2ϕ(h/2)-ϕ(h)$. The solution makes sense if i check it but how do i come to that solution? And do i use the same logic for part a or is there something different with a? (because the power is different.)
If $$\phi(h)=L-c_1h-c_2h^2-c_3h^3-\cdots$$ then $$\phi(h/2)=L-(1/2)c_1h-(1/4)c_2h^2-(1/8)c_3h^3-\cdots$$ so we see the term in $h$ will cancel if we compute $$2\phi(h/2)-\phi(h)=L+(1/2)c_2h^2+(3/4)c_3h^3+\cdots$$ and give us that $L$ is estimated by $2\phi(h/2)-\phi(h)$ with error on the order of $h^2$.
Now see what you can do with (a).