CT3 CMP Upgrade 11

CT3: CMP Upgrade 2010/11
Page 1
Subject CT3
CMP Upgrade 2010/11
CMP Upgrade
This CMP Upgrade lists all significant changes to the Core Reading and the ActEd
material since last year so that you can manually amend your 2010 study material to
make it suitable for study for the 2011 exams. It includes replacement pages and
additional pages where appropriate. Alternatively, you can buy a full replacement set of
up-to-date Course Notes at a significantly reduced price if you have previously bought
the full price Course Notes in this subject. Please see our 2011 Student Brochure for
more details.
This CMP Upgrade contains:
All changes to the Syllabus objectives and Core Reading.
Changes to the ActEd Course Notes, Series X Assignments and Question and
Answer Bank that will make them suitable for study for the 2011 exams.
The Actuarial Education Company
IFE: 2011 Examinations
Page 2
Changes to the Syllabus objectives and Core Reading
1.1
Syllabus objectives
No changes have been made to the syllabus objectives.
1.2
Core Reading
No changes have been to the Core Reading.
Page 3
Changes to the ActEd Course Notes

Chapter 12
Solutions 12.1 and 12.2 have been rewritten. Replacement pages are provided at the
end of this document.
Chapter 14
Section 2.2 has been rewritten. Replacement pages are provided at the end of this
document.
Page 4
Changes to the Q&A Bank

There have been no significant changes to the Q&A Bank.
Page 5
Changes to the X Assignments

Assignment X3
Solution X3.6(ii)
This part has been rewritten as follows:
(ii)
Confidence interval for the probability that the target plane is destroyed
The probability a that a target plane is destroyed is given by a = 1 - (1 - p )3 .
[1]
So a 90% confidence interval for a can be obtained by inserting the limits on the 90%
confidence interval for p obtained in (i) into the formula for a given above.
Hence, the required confidence interval is:
(1 - (1 - 0.646) , 1 - (1 - 0.794) ) = (0.956, 0.991)

2
[1]
[Total 2]
Assignment X4
Solution X4.10(iii)
This part of the solution has been rewritten as follows:

(iii)
Comment
The paired test shows that there was a significant difference between the two
dessicators, whereas the two-sample test does not indicate any significant difference. [1]
The small but significant difference between the two desiccants is masked in the twosample test because the test statistic for the two-sample test is calculated using the
pooled variance (which is 0.1657) rather than the sample variance of the differenced
data (which is 0.01411). A smaller variance leads to a larger test statistic, which means
we are more likely to reject the null hypothesis. In other words, the increased power of
the paired test enables a significant difference to be identified.
[1]
[Total 2]
Page 6
Solution X4.11
In part (ii), the following statements of the null and alternative hypotheses have been
added:
H 0 : the mean is the same for each category
H1 : the means are not the same for all of the categories
The inequalities halfway down Page 16 are the wrong way round. This line should
read:
y1i > y2i > y3i > y4i
Page 7
Other tuition services

In addition to this CMP Upgrade you might find the following services helpful with
your study.
5.1
Study material
We offer the following study material in Subject CT3:
Series Y Assignments
ASET (ActEd Solutions with Exam Technique) and Mini-ASET
Flashcards
Mock Exam A and Mock Exam B (based on Mock Exam 2010 and Mock Exam
2009, respectively).
Revision Notes.
For further details on ActEds study materials, please refer to the 2011 Student
Brochure, which is available from the ActEd website at www.ActEd.co.uk.
5.2
Tutorials
We offer the following tutorials in Subject CT3:
a set of Regular Tutorials (lasting two or three full days)
a Block Tutorial (lasting two or three full days)
a Revision Day (lasting one full day).
For further details on ActEds tutorials, please refer to our latest Tuition Bulletin, which
is available from the ActEd website at www.ActEd.co.uk.
5.3
Marking
You can have your attempts at any of our assignments or mock exams marked by
ActEd. When marking your scripts, we aim to provide specific advice to improve your
chances of success in the exam and to return your scripts as quickly as possible.
For further details on ActEds marking services, please refer to the 2011 Student
Brochure, which is available from the ActEd website at www.ActEd.co.uk.
Page 8
Feedback on the study material

ActEd is always pleased to get feedback from students about any aspect of our study
programmes. Please let us know if you have any specific comments (eg about certain
sections of the notes or particular questions) or general suggestions about how we can
improve the study material. We will incorporate as many of your suggestions as we can
when we update the course material each year.
If you have any comments on this course please send them by email to CT3@bpp.com
or by fax to 01235 550085.
CT3-12: Hypothesis testing
Page 41
Chapter 12 Solutions
Solution 12.1
(i)
The probability of a Type I error is given by:

P (reject H 0 when H 0 true) = P ( X > 28 when X ~ Exp (1/ 20))
= 1 - FX (28) = e -28/ 20 = 0.2466
The CDF of the exponential distribution is given on Page 12 of the Tables.

(ii)
The probability of a Type II error is given by:

P (do not reject H 0 when H 0 false) = P ( X < 28 when X ~ Exp (1/ 30))
= FX (28) = 1 - e -28/ 30 = 0.6068
In this case we were forced to choose between H 0 : m = 20 and H1 : m = 30 .

So saying that H 0 is false is the same as saying that m = 30 .
Since weve only got one value in our sample here, not surprisingly, the
probabilities of Type I and Type II errors are quite big.
Solution 12.2
The power is the probability of rejecting H 0 when the true value of the parameter is
some value other than = 20 . In terms of m this is:
P ( X > 28 | X ~ Exp (1/ m )) = 1 - FX (28) = e -28/ m
If m is large (1,000, say), then the power will be close to 1, since the test will reject
H 0 : m = 20 very easily. Conversely if m is small (10, say), then the power will be
close to 0, since the test will not reject H 0 : m = 20 very easily.
Page 42
CT3-12: Hypothesis testing
Solution 12.3
Given a single value from an exponential distribution, the Neyman-Pearson criterion is

L
reject H 0 if 0 < critical value . Using the null and alternative hypotheses given in
L1
Question 12.1, the test becomes:

1 20x
e
20
< constant
1 30x
e
30
This reduces to e 60 < constant , or x > constant . This was exactly the form of the
test that we used (we rejected H 0 when x > 28 ). So this is a Neyman-Pearson test.
x
Solution 12.4
(i)
We want:
0.05 =
k
1 - 20x
-x
-k
e dx = - e 20 = e 20
20
So:
k = 20 ln 0.05 = 59.9
(ii)
The probability of a type II error is:

k
k
1 - 30x
- e - 30x = 1 - e -1.997 = 0.864
e
dx
=
30
0
0
CT3-14: Analysis of variance
Page 19
Examining the treatment means

In this section we will be finding confidence intervals for a single treatment mean or
between a pair of treatment means using the results of Chapter 11. However, we will be
SS R
using our estimate for the overall common variance, s 2 =
, rather than s 2 or sP2 .
n-k
2.1
Confidence intervals for a single treatment mean

In the situation where interest is focused on a particular treatment, say treatment
i, 2 can be estimated using the residual mean square 2 and a confidence
interval for + i (ie for treatment mean mi ) is given by:
y i i ts
ni
where t is based on (n k ) degrees of freedom.
Recall from Chapter 11, Section 3.1 that a confidence interval for the mean from a
normal distribution was based on:
X -m
~ tn -1
S n
We are going to use this to obtain a confidence interval for treatment mean i, mi , using
our estimate yi i . We will use s 2 instead of s 2 , so the t-distribution will have n - k
degrees of freedom:
yi i - mi
~ tn - k
s ni
where s 2 =
SS R
.
n-k
This rearranges to the confidence interval given in the Core Reading above.
Page 20
2.2
CT3-14 Analysis of variance
Confidence intervals for a pair of treatment means

In the situation where interest is focused on a particular pair of treatments, say
treatments 1 and 2 for convenience, then:
1
1
var Y1i - Y2i = s 2
+
n1 n2
and a confidence interval for m1 - m2 = ( m + t 1 ) - ( m + t 2 ) = 1 2 is given by:

1
1
( y 1i - y 2i ) ts n + n
1
2
where t is again based on (n k ) degrees of freedom.
Recall from Chapter 11, Section 5.1 that a confidence interval for the difference
between two means from normal distributions was based on:
( X1 - X 2 ) - ( m1 - m2 ) ~ t
SP
1
n1
1
n2
n1 + n2 - 2
This result can be adapted to obtain a confidence interval for the difference between
treatment means 1 and 2, m1 - m2 , using our estimate y1i - y2i . We will use s 2 , which
is a generalisation of the pooled variance to the k -sample case. We have already seen
that:
S P2
n1 - 1) S12 + (n2 - 1) S 22 (n1 - 1) S12 + (n2 - 1) S 22

(
=
=
(n1 - 1) + (n2 - 1)
n1 + n2 - 2
We now extend this to define:
s 2 =
(n1 - 1) S12 + (n2 - 1) S22 + (nk - 1) Sk2

(n1 - 1) + (n2 - 1) + + (nk - 1)
n1 - 1) S12 + (n2 - 1) S22 + (nk - 1) Sk2

(
=
n-k
Page 21
Note that the numerator of this expression is the residual sum of squares, SS R , and that:
(n - k )s 2
s
~ c n2- k
Using the definition of the t distribution (from Chapter 9), it follows that:
(Y1i - Y2i ) - ( m1 - m2 ) ~ t
s
1
n1
1
n2
n-k
and from this result we get the confidence interval given in Core Reading above.
Question 14.14
Using the pulse rates from Question 14.1

Sales
Accounts
Admin
72
65
61
81
73
79
96
78
67
75
66
74
69
88
74
102
65
In Question 14.2, we found that y1i = 82.125, y2i = 70.5, y3i = 69.2 . In Question 14.10,
we found that SST = 1,986.12, SS B = 649.44, SS R = 1,336.68 .
(i)
Show using ANOVA that there is no significant difference between the

treatment means.
(ii)
(a)
Show that the 95% confidence interval for the difference between the
mean pulse rates for Sales and Admin is (0.976, 24.9)
(b)
What does this confidence interval mean?
(c)
Explain the apparent contradiction between the confidence interval and

the ANOVA result.
Page 22
2.3
Analysing treatment means using a least significant difference

approach
Constructing such intervals for all possible pairs of treatments is not
recommended the interpretation of them becomes difficult as the overall level
of confidence of the whole set of intervals has to be considered. However, if
H 0 : i = 0, i = 1, 2,..., k has been rejected, a good idea as to whether the
treatments fall into several reasonably homogeneous groups can be obtained as
follows.
So if there is a significant difference between the treatment means we can compare

treatment means to see which are the same and which are different.
Step 1
List the observed treatments in order, eg with k = 4 we might have:
y 2i < y 3i < y 1i < y 4i
Step 2
We will now examine each of the pairs in order to see whether the means are the same
or not. We do this by using a two-sample test. For example on the first pair:
H 0 : m2 = m3
H1 : m2 m3
Note that Core Reading carries out a two-sided test and this is what is expected in the
exam. It would be natural to choose a one-sided test (to see if m3 > m2 ) as we have
arranged the means in increasing order, however, this is not the approach used.
Our statistic is:
(Y3i - Y2i ) - ( m3 - m2 ) ~ t
s
1
n3
1
n2
n-k
Page 23
Now, under H 0 , m3 - m2 = 0 . Since y3i > y2i there will be a significant difference
between the means if the statistic is greater than the upper 2% critical value of the
appropriate t distribution:
( y3i - y2i ) - ( m3 - m2 ) > t

s
1
n3
1
n2
0.025,n - k
Rearranging:
( y3i - y2i ) > t0.025,n - k s
1
n3
1
n2
The value on the right hand side is the least significant difference, ie the value that the
difference between the sample means needs to exceed to say that there is a significant
difference.
If the ni s are the same for each of the treatments, then the least significant differences
will be the same. This makes it a quick way to test each of the pairs of means. A
standard two-sample test is also fine.
For a given level of significance, say 5%, calculate the least difference between
y 3i and y 2i which would be significant, namely:
1
1
where t = t0.025,n - k
ts
+
n2 n3
ie the value of a t n k variable which is exceeded with probability 0.025. If the
difference y 3i - y 2i is less than this least significant difference then it can be
indicated that the treatment means fall into the same group, for example by
underlining the pair. This process can be repeated for y 3i and y 1i and then for
y 1i and y 4 i . As an example the results may give:
y 2i < y 3i < y 1i < y 4 i
This indicates that treatment 4 is on its own.
Page 24
Since y2i , y3i fall into the same group and y3i , y1i fall into the same group, it is worth
checking to see if y2i and y1i fall into the same group. If they were this would mean
all three of these means fall into the same group and we would show this as:
y2i < y3i < y1i < y4i
Core reading now analyses its example from earlier:

Company 1: 36, 28, 32, 43, 30, 21, 33, 37, 26, 34
n1 = 10
y1i = 32
Company 2: 26, 21, 31, 29, 27, 35, 23, 33
n2 = 8
y2i = 28.125
Company 3: 39, 28, 45, 37, 21, 49, 34, 38, 44
n3 = 9
y3i = 37.2
SS R = 1,114.43 s 2 =
SS R 1,114.43
=
= 46.43
n-k
24
Step 1
In the example above, y 2i < y 1i < y 3i .
Step 2
For a 5% level, the least significant difference between y 1i and y 2i is:
1 1
2.064 46.43
+
10 8
= 6.67
Using the fact that t0.025,24 = 2.064 .

y 1i - y 2i = 3.9 (1 dp) so y 2i < y 1i
For y 3i and y 1i the least significant difference is 6.46 and y 3i - y 1i = 5.2 (1 dp)
so:
y 2i < y 1i < y 3i
Page 25
Question 14.15
(i)
Show that the least significant difference between y3i and y1i is 6.46.
(ii)
By calculating the least significant difference between y2i and y3i , confirm that
we should not underline all three treatments.
Note that this result does not contradict the earlier rejection of H0 . If in advance
it had been decided to concentrate on companies 2 and 3 (while at the same time
using the information from company 1) a 95% confidence interval for t 3 - t 2
would be obtained as:
ie
(37.2 - 28.1) 2.064
1 1
46.43 +
9 8
9.1 6.8
(2.3,15.9)
ie
and this interval does not contain zero.
So there is a significant difference between the means of companies 2 and 3. However,

just because the biggest and smallest company means are different from each other
doesnt necessarily imply that all of the treatment means are different from the overall
mean.
Question 14.16
The bowling ball scores in Question 14.6 were:

Ball
1
2
3
Score
173, 166, 179, 183, 199
195, 206, 188, 195, 210, 221
169, 155, 172, 188
sum
900
1,215
684
sum of squares
162,616
246,771
117,514
Where SS R = 1,899.5 .
Analyse the means of the bowling ball scores in Question 14.6 using a least significant
difference approach.
Page 26
Further comments
ANOVA for a completely randomized comparative experiment on only k = 2
treatments is equivalent to the two-sample t test of Chapter 12. To verify this,
one must be aware that t d2 F1,d .
It is no coincidence that the sums of squares can be split up in the same way for
ANOVA and linear regression. In fact linear regression is just a special type of
ANOVA and the results of a regression analysis can be presented in an ANOVA table.
A linear regression analysis of:
Yi = a + bx i + e i , i = 1, 2,..., n
can be represented in the ANOVA framework as follows:
(y i - y ) = (y i - y i ) + (y i - y )
2
SST = SSR + SSREG
Slightly different abbreviations have been used here for the sum of squares.
Source of variation
Degrees of
Freedom
Sums of
Squares
Mean Squares
Regression
Residual
1
n-2
SSREG
SSREG
Total
n -1
SST
Under H0 : b = 0,
SSR
SSR / (n - 2)
SSREG
is F1,n - 2 .
SSR / (n - 2)
A one-way ANOVA analysis of k treatments is equivalent to a regression in which

the response is regressed on (k 1) dummy explanatory variables, each of
which takes values 0 or 1 only.

CT3 CMP Upgrade 11

Caricato da

Informazioni sul documento

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

CT3 CMP Upgrade 11

Caricato da

Copyright:

Formati disponibili

CT3: CMP Upgrade 2010/11

This CMP Upgrade contains:

All changes to the Syllabus objectives and Core Reading.

The Actuarial Education Company

IFE: 2011 Examinations

CT3: CMP Upgrade 2010/11

Changes to the Syllabus objectives and Core Reading

IFE: 2011 Examinations

The Actuarial Education Company

CT3: CMP Upgrade 2010/11

Changes to the ActEd Course Notes

The Actuarial Education Company

IFE: 2011 Examinations

CT3: CMP Upgrade 2010/11

Changes to the Q&A Bank

IFE: 2011 Examinations

The Actuarial Education Company

CT3: CMP Upgrade 2010/11

Changes to the X Assignments

The probability a that a target plane is destroyed is given by a = 1 - (1 - p )3 .

(1 - (1 - 0.646) , 1 - (1 - 0.794) ) = (0.956, 0.991)

This part of the solution has been rewritten as follows:

The Actuarial Education Company

IFE: 2011 Examinations

CT3: CMP Upgrade 2010/11

IFE: 2011 Examinations

The Actuarial Education Company

CT3: CMP Upgrade 2010/11

Other tuition services

ASET (ActEd Solutions with Exam Technique) and Mini-ASET

a set of Regular Tutorials (lasting two or three full days)

a Block Tutorial (lasting two or three full days)

a Revision Day (lasting one full day).

The Actuarial Education Company

IFE: 2011 Examinations

CT3: CMP Upgrade 2010/11

Feedback on the study material

IFE: 2011 Examinations

The Actuarial Education Company

CT3-12: Hypothesis testing

The probability of a Type I error is given by:

The CDF of the exponential distribution is given on Page 12 of the Tables.

The probability of a Type II error is given by:

In this case we were forced to choose between H 0 : m = 20 and H1 : m = 30 .

The Actuarial Education Company

IFE: 2011 Examinations

CT3-12: Hypothesis testing

Given a single value from an exponential distribution, the Neyman-Pearson criterion is

Question 12.1, the test becomes:

The probability of a type II error is:

IFE: 2011 Examinations

The Actuarial Education Company

CT3-14: Analysis of variance

Examining the treatment means

Confidence intervals for a single treatment mean

where t is based on (n k ) degrees of freedom.

The Actuarial Education Company

IFE: 2011 Examinations

CT3-14 Analysis of variance

Confidence intervals for a pair of treatment means

and a confidence interval for m1 - m2 = ( m + t 1 ) - ( m + t 2 ) = 1 2 is given by:

where t is again based on (n k ) degrees of freedom.

n1 - 1) S12 + (n2 - 1) S 22 (n1 - 1) S12 + (n2 - 1) S 22

We now extend this to define: