Академический Документы
Профессиональный Документы
Культура Документы
1 (i)
(ii)
x
f(x)
q(x)
2623/01
1
0
0
2
-4
-4
3
-3
-3
4
???
3
June 2004
5
8
14
x
f(x)
()
2()
3()
1
0
3
-3
-3
5
8
11
14
7
108
100
89
75
p(x) = 0 - 3(x-1)/2 + 14(x-1)(x-3)/(22.2!) + 75(x-1)(x-3)(x-5)/(23.3!)
p(4) =
-3.9375
verify:
[M1A1]
estimate:
[A1]
[subtotal 3]
other methods
may be used
[M1A2]
[M1A2]
[A1]
[subtotal 7]
(i) x = (x+k/x)/2
2x2 = x2 + k
x2 = k
Root found has same sign as initial value.
[M1A1]
[A1]
[E1]
[subtotal 4]
f '(x) = 2x
(ii) f(x) = x2 k
NR: xr+1 = xr - (xr2 - k)/(2xr) = xr - xr/2 +
k/(2xr)
= xr/2 + k/(2xr) = (xr + k/xr) / 2
[B1]
[M1A1]
[A1]
[subtotal 4]
f '(x) = 3x2
(iii) f(x) = x3 - k
NR: xr+1 = xr - (xr3 - k)/(3xr2)
= (2xr + k/xr2) / 3
r
xr
0
1
[B1]
[M1A1]
(not required)
1
2
3
4
5
3.3333333 2.4622222 2.0813412 2.0031374 2.0000049
(6)
(2)
[M1]
[A3]
[-1 each error]
[subtotal 7]
[TOTAL 15]
3 (i)
2623/01
June 2004
[M1A1]
g(1)2
0.018
1.1066666
1.1166666
1.0966666
1
1.5
0.5
1
[B1B1B1]
[subtotal 5]
[M1]
[A1]
[A1]
[A1]
[subtotal 4]
[M1]
[A1]
[A1]
[A1]
[subtotal 4]
[B1]
[E1]
[subtotal 2]
[TOTAL 15]
(i) M1 = sqrt(cos(0.5)) =
M2 = (sqrt(cos(0.25)) + sqrt(cos(0.75)))/2 =
M4 = (sqrt(cos(0.125)) + )/4 =
M8 = (sqrt(cos(0.0625)) + )/8 =
(ii) M1
M2
M4
M8
0.9367937
0.9198608
0.9154692
0.9143569
0.9367937
0.9198608
0.9154692
0.9143569
diffs
ratio
-0.0169329
-0.0043915 0.2593483
-0.0011123 0.2532868
differences reducing by a factor of 4 at each step
process is of order h2
[M1A1]
[M1A1]
[M1A1]
[M1A1]
[subtotal 8]
[M1A1]
[E1]
[E1]
[subtotal 4]
[M1]
[A1]
[B1]
[subtotal 3]
[TOTAL 15]
Examiners Report
In part (i), candidates generally had little difficulty verifying the given results and
estimating f(4). A number chose to obtain the quadratic from the data rather than
check the data fitted the polynomial. This is a perfectly correct approach, but
somewhat more work than was intended.
Part (ii) attracted many good attempts, but there were quite a few elementary errors.
Amongst the most common were taking x0 as zero and taking h as 1.
In parts (iii) and (iv) there was some rather vague discussion offered in hope rather
than expectation. The essential points were to do with interpolation and extrapolation
in part (iii), and some consideration of using all the available data in part (iv).
Q.2
Very large numbers of candidates were unable to tackle the algebra in this question.
In part (i) the requirement was to reduce x = (x + k/x) to x2 = k, but this proved
beyond many. Even fewer could derive the Newton-Raphson formula in part (ii), the
usual error being to write (x2 k) as x2 k. Surely AS/A2 candidates should be
doing better than this.
In part (iii) some did very well, but others were totally adrift, apparently unable to see
the simple link with previous parts.
It was surprising that so few did this question well as it might be thought an obvious
teaching topic when dealing with the Newton-Raphson method.
Q.3
The obvious mistake, using the wrong denominator for maximum possible relative
error, was prevalent right from the start in this question.
In parts (i) and (ii) most candidates chose to recalculate everything at every stage
rather than to rely on knowledge and efficient methods. In part (i), it was enough to
quote results regarding relative error in reciprocals, squares and square roots. In
part (ii), it was quite easy to find the error in Simpsons rule without recalculating the
integral.
In part (iii) the estimate of g(1) and hence the true value lies between 0.5 and
1.5. In part (iv) the estimate comes to 1. It is therefore a little weak to suggest that
the latter might be wrong.
Q.4
Most, but by no means all, were able to evaluate the mid-point rule estimates in part
(i). In part (ii) the requirement was to look at the differences and the ratios of
differences. The fact that the ratios are about indicates that the mid-point rule has
second order convergence. Many candidates seemed to know this but expressed
themselves poorly.
In part (iii) the extrapolation and the rounding (to 3 s.f.) was usually done well.
1