where, as x → 0 the error [ = RN(x)] is much much
smaller than xN - the last and smallest term which has been
included in PN.
Thus, fixing N, PN(x) is a very good approximation for x
near 0.
The global context is that for all x in an interval,
f(x) = PN(x) + error,
where, as N → ∞, the error [ = RN(x)] is
small - the infinite series ∑n=0∞ an xn
converges to the value f(x).
Thus, fixing x [or varying x over a fixed interval], PN(x)
is a very good approximation for N very large.
From the local viewpoint, since
P0(x) = f(0),
the statement f(x) ≈ P0(x) gives the statement f(x) ≈ f(0) for x near 0 - or f(x) is continuous at
x=0.
Note that
P1(x) = f(0) + f′(0)·(x − 0)
gives the tangent line approximation to f(x).
The global viewpoint is required to give an efficient way
to calculate, e. g., ex, for, say −0.1 ≤ x ≤ 0.1
For example, in treating continuous compounding of interest, we
might need er, with r=6.1% = .061 or r=6.9% = .069. A
very good approximation is
er ≈ 1 + r +
r2
2
, −0.1 ≤ x ≤ 0.1.
File translated from
TEX
by
TTH,
version 4.03. On 08 Feb 2012, 18:33.