Split exam: Applied Statistics

Lecturer: Prof. Visieu Lac

Mock Exam January/2018

Name:

Registration No.:

__________________________
Signature Student
Directions:

• The exam is worth 60 points and with 30 points you have passed the
exam.
• The following means are allowed:
o Non programmable calculator
o Formula sheet (provided with the exam)
o Statistical tables (provided with the exam)
The usage of any other means than the ones listed above is seen as
cheating!
• Check the exam for completeness. It consists of 5 exercises and 13
pages.
• You may not tear out or remove any pages from the exam paper, or
remove the staple
• Please write neatly and show sufficient working to make your methods
and results clear. Answers without working may gain no points.
• Where appropriate, round your answers to 4 decimal places.

Good luck

Result in total points:

_______________________________
Signature Lecturer

Page 1 of 13
 Split exam: Applied Statistics
Lecturer: Prof. Visieu Lac

Question 1: Case Study (12 points)

The manager of a logistic company believes that the average petrol price
has increased by more than 20% from previous months. If true, this
would hurt the company’s operating margin. The manager asks you to
verify his claim.

In your initial research on petrol prices, you find that an industry body has
recently conducted a large-scale survey on petrol prices for the previous
months. They found that the average price was €1.30 and a standard
deviation of €0.16.

Take the large-scale survey on petrol prices to be the population.

a) Assume that the distribution of the petrol prices is normal, please

calculate the probability that petrol price is at least €1.56. (3 points)

To determine if there is sufficient evidence to support the manager’s

claim, you conduct a survey of your own based on the current economic
conditions.

b) Formulate the null and alternative hypotheses to test if the

manager’s claim is correct. (2 points)

Page 2 of 13
 Split exam: Applied Statistics
Lecturer: Prof. Visieu Lac

c) Discuss two techniques of selecting random samples that are

appropriate for your survey. (3 points)

d) After conducting a short survey, you find that your sample of 16

observations has a mean of €1.65 and standard deviation of €0.27.
Discuss the importance of taking the standard deviation into
consideration. (3 points)

e) Based on your sample, calculate the test statistics and p-value. (5

points)

Page 3 of 13
 Split exam: Applied Statistics
Lecturer: Prof. Visieu Lac

f) At α = 0.05, conclude whether there is sufficient evidence to support

the manager’s claim. (2 points)

Question 2: Simple Regression Model (20 points)

An important application of regression analysis in accounting is in the

estimation of cost. By collecting data on volume and cost and using the
least squares method to develop an estimated regression equation
relating volume and cost, an accountant can estimate the cost associated
with a particular manufacturing volume. Consider the following sample of
production volumes and total cost data for a manufacturing operation.

Production volume (units) Total cost (€)

400 4000
450 5000
550 5400
600 5900
700 6400

a) Use these data to develop an estimated regression equation that

could be used to predict the total cost for a given production volume.
(10 points)

Answer:
Production volume (X) Total cost (Y) X - XBar Y - YBar (3) x (4) (3) x (3)
(1) (2) (3) (4)
400 4,000 -140.00 -1,340.00 187,600.00 19,600.00
450 5,000 -90.00 -340.00 30,600.00 8,100.00
550 5,400 10.00 60.00 600.00 100.00
600 5,900 60.00 560.00 33,600.00 3,600.00
700 6,400 160.00 1,060.00 169,600.00 25,600.00
Mean = 540.00 5,340.00 Sum = 422,000.00 57,000.00
(1)' (2)' (5)' (6)'

b1 = (5)’/(6)’ = 422,000 / 57,000 = 7.4035

b0 = Y − b1 X = 5340 − 7.4035× 540 =1342.11

Regression model: Y = β0 + β1 X + ε
The estimated regression equation: ŷ = b0 + b1 x =1, 342.11+ 7.4035x
Page 4 of 13
 Split exam: Applied Statistics
Lecturer: Prof. Visieu Lac

b) What is the variable cost per unit in your model? (2 points)

Answer:
The variable cost per unit is equal to b1 i.e. €7.40. (Note: Total cost = Fixed
costs + Variable costs. By definition, fixed costs are independent of the number
of units i.e. constant whereas variable costs vary with the number of units x.)

c) Compute the coefficient of determination. What percentage of the

variation in total cost can be explained by the variation in production
volume? (8 points)

Answer:
The coefficient of determination is

RSS
R2 = 1 −
TSS

where
n
RSS = ∑ ( y − yˆ )
i =1
i i
2

n
TSS = ∑ ( y − y)
i =1
i
2

Production volume (X) Total cost (Y) X - XBar Y - YBar Predicted Profit [(2) - (7)]^2 [(4)]^2 [(2)' - (7)]^2
(1) (2) (3) (4) (7)
400 4,000 -140.00 -1,340.00 4,303.509 92,117.575 1,795,600 1,074,314.066
450 5,000 -90.00 -340.00 4,673.684 106,481.994 115,600 443,976.731
550 5,400 10.00 60.00 5,414.035 196.984 3,600 5,481.194
600 5,900 60.00 560.00 5,784.211 13,407.202 313,600 197,322.992
700 6,400 160.00 1,060.00 6,524.561 15,515.543 1,123,600 1,403,185.719
540.00 5,340.00 Sum = 227,719.298 3,352,000.00 3,124,280.702
(1)' (2)' (8) (9) (10)
RSS TSS ESS

RSS (8) = 227,719.298

TSS (9) = 3,352,000.000

Thus

RSS 227, 719.298

R2 = 1− = 1− = 0.9321 or 93%
TSS 3, 352, 000

Page 5 of 13
 Split exam: Applied Statistics
Lecturer: Prof. Visieu Lac

A coefficient of determination of 0.93 means that 93% of the variations in the

total cost can be explained by the variations in the production volume.

Alternatively, we can calculate R2 by noting that in simple regression R is equal

to the Pearson correlation coefficient r (or ρ) where

r=
σ XY
=
∑(x − x)(y − y)
i i

σ Xσ Y ∑(x − x) ∑(y − y)
i
2
i
2

Note that σ σ have already been calculated from the table in a) (i.e.
(5)’ of column 5, and square root of (6)’ of column 6 respectively). All that
remains to be done is to calculate σY. This gives σY =square root of 3,352,000.

 σ XY   
2 2
422, 000
 =  = 0.9321
2 2
R =r =
 σ Xσ Y   57, 000 × 3352000 

This method in fact involves fewer computations because one does not have to
calculate the quantities related to the predicted profits (columns 8 and 10).

Question 3: Multiple Regression Model (20 points)

A microcomputer manufacturer has developed a regression model relating

his sales (Y in €10,000s) with three independent variables. The three
independent variables are price per unit (PRICE in €100s), advertising
(ADV in €1,000s) and the number of product lines (LINES). Part of the
regression results is shown below.

Coefficient Standard Error

Intercept 1.0211 22.8752
PRICE –0.1524 0.1411
ADV 0.8849 0.2886
LINES –0.1463 1.5340

Source of Variation Degrees of Freedom Sum of Squares

Regression 2708.61
Residual (or Error) 2840.51

Page 6 of 13
 Split exam: Applied Statistics
Lecturer: Prof. Visieu Lac

a) Please use the above results and write the regression equation
that can be used to predict sales. (2 points)

Answer:
The regression equation is:

Ŷ =1.0211- 0.1524 * PRICE + 0.8849 * ADV - 0.1463 * LINES

b) If the manufacturer has 10 product lines, advertising of €40,000, and

the price per unit is €3,000, what is your estimate of their sales?
Please give your answer in euros. (3 points)

Answer:
Substituting the values gives:

Ŷ = 1.0211- 0.1524 * 30 + 0.8849 * 40 - 0.1463 * 10

=30.3821 (in ten thousand euros)

Since Y is measured in 10,000, the estimate sales in euros is €303,821.

c) Interpret the coefficient LINES. (2 points)

Answer:
Every increase in the number of product lines will reduce the sales by €1,463
while keeping all other variables constant.

d) If the size of the sample is 18, at α = 0.05, test to see if there is a

significant relationship between sales and the number of product
lines. Be sure to state the null and alternative hypotheses. (8 points)

Answer:
H0 : β 3 = 0 There is no relationship between Sales and product lines
H1 : β 3 ≠ 0 There is a linear relationship between Sales and product lines

Test statistics:

b1 − β3 −0.1463− 0
t= = = −0.0954
sb3 1.5340

Page 7 of 13
 Split exam: Applied Statistics
Lecturer: Prof. Visieu Lac

Looking up the probability from the t distribution table with n-k-1 = 14 degrees
of freedom gives a probability greater than 0.2. The p-value is twice that
probability (two-tail) thus p-value is greater than 0.4. This is greater than 0.05
(the assumed level of significance). Therefore we retain H0 i.e. at a level of
significance of 5%, there is insufficient evidence to show that there is linear
relationship between Sales and the number of product lines.

e) Calculate the coefficient of determination and fully interpret its

meaning. (5 points)

Answer:
First we complete the table:

Source of Variation Degrees of Freedom Sum of Squares

Regression 2708.61 = ESS
Residual (or Error) 2840.51
Total 5549.12 = TSS
(=2708.61+2840.51)

The coefficient of determination is given by

Explained Sum of Squares (ESS) 2708.61
R2 = = = 0.488
Total Sum of Square (TSS) 5549.12

This means that 48.8% of the variations in sales can be explained by

the variations in price, advertising and number of product lines.

Page 8 of 13
 Split exam: Applied Statistics
Lecturer: Prof. Visieu Lac

Question 4: Logistic Regression (10 points)

A study designed to assess the association between obesity (defined as

BMI > 30) and incident cardiovascular disease. Data were collected from
participants who were between the ages of 35 and 65, and free of
cardiovascular disease (CVD) as baseline.

The logistic regression analysis reveals the following (1=subjects with

CVD; 0=Otherwise. Obesity is an indicator variable in the model, coded as
follows: 1=obese and 0=not obese):

Regression
Variable Chi-square p-value
Coefficient
Intercept -2.367 307.38 0.0001
Obesity 0.658 9.87 0.0017

a) Write down the logistic regression model. (2 points)

Logistic regression model:

b) Calculate the odds ratio of developing CVD among obese persons as

compared to non-obese persons. Interpret the result. (5 points)

Odds ratio

Interpretation:
The odds ratio is larger than 1, which means that knowing a person
is obese positively influences the probability that this person
develops CVD.

More precisely, the odds in favour of developing CVD in case of

obesity is 93% larger compared to the odds in favour of developing
CVD in case of non-obesity.

c) Calculate the probability of developing CVD given that the person is

not obese. (3 points)

X = 0 (person is not obese)

Page 9 of 13
 Split exam: Applied Statistics
Lecturer: Prof. Visieu Lac

Question 5: Forecasting (30 points)

The following time series gives the number of units sold during 5 years at
a boat dealership.

Year Quarter Number of Units

1 1 300
2 240
3 240
4 290
2 1 350
2 300
3 280
4 320
3 1 410
2 400
3 390
4 410
4 1 490
2 450
3 440
4 510

a) Find the four-quarter centred moving averages. (4 points)

N u m be r of Un it s MA Ce nt r e d M A
Ye a r Qua r t e r
( Y) ( Pe r iod 4 ) ( CM A)

1 300
2 240 267.5
1
3 240 280 273.75
4 290 295 287.5
1 350 305 300
2 300 312.5 308.75
2
3 280 327.5 320
4 320 352.5 340
1 410 380 366.25
2 400 402.5 391.25
3
3 390 422.5 412.5
4 410 435 428.75
1 490 447.5 441.25
2 450 472.5 460
4
3 440
4 510

Page 10 of 13
 Split exam: Applied Statistics
Lecturer: Prof. Visieu Lac

b) Compute the seasonal-irregular component. (3 points)

N u m be r of Un it s MA Ce nt r e d M A
Ye a r Qua r t e r S t x I t = Y/ CM A
( Y) ( Pe r iod 4 ) ( CM A)

1 300
2 240 267.5
1
3 240 280 273.75 0.8767
4 290 295 287.5 1.0087
1 350 305 300 1.1667
2 300 312.5 308.75 0.9717
2
3 280 327.5 320 0.8750
4 320 352.5 340 0.9412
1 410 380 366.25 1.1195
2 400 402.5 391.25 1.0224
3
3 390 422.5 412.5 0.9455
4 410 435 428.75 0.9563
1 490 447.5 441.25 1.1105
2 450 472.5 460 0.9783
4
3 440
4 510

c) Compute the seasonal factors for all four quarters. (4 points)

Yea r Qua r t er S t x I t = Y/ CM A Q1 Q2 Q3 Q4

1
2
1
3 0.8767 0.8767
4 1.0087 1.0087
1 1.1667 1.1667
2 0.9717 0.9717
2
3 0.8750 0.8750
4 0.9412 0.9412
1 1.1195 1.1195
2 1.0224 1.0224
3
3 0.9455 0.9455
4 0.9563 0.9563
1 1.1105 1.1105
2 0.9783 0.9783
4
3
4
Se
S asona l fa ct or s S t 1.1322 0.9908 0.8991 0.9687

Adjusted
N or m alise d se
s a sona l fa ct or s S t 1 .1 3 4 8 0 .9 9 3 1 0 .9 0 1 1 0 .9 7 1 0

Note: The adjusted seasonal factors ensure that the average of all the seasonal
factors adds up to 1.

Page 11 of 13
 Split exam: Applied Statistics
Lecturer: Prof. Visieu Lac

d) Compute the deseasonalized time series for sales. (4 points)

The deseasonlised time series is obtained by dividing the actual observations

(Y) by the adjusted seasonal factors. (Note that there may be slight
discrepancies in the fifth digit onwards because of rounding.)

D e se a son a le d
N u m be r of Un it s
Ye a r Qu a r t er S t x I t = Y/ CM A Q1 Q2 Q3 Q4 Tim e Se r ie s
( Y)
( 1)
1 300 264.357
300/1.1348
2 240 241.677
1
3 240 0.8767 0.8767 266.328
4 290 1.0087 1.0087 298.672
1 350 1.1667 1.1667 308.416
2 300 0.9717 0.9717 302.096
2
3 280 0.8750 0.8750 310.716
280/0.9011
4 320 0.9412 0.9412 329.570
1 410 1.1195 1.1195 361.288
2 400 1.0224 1.0224 402.794
3
3 390 0.9455 0.9455 432.783
4 410 0.9563 0.9563 422.261
1 490 1.1105 1.1105 431.783
2 450 0.9783 0.9783 453.144
4
3 440 488.268
4 510 525.252
S
Sea son a l fa ct or s S t 1.1322 0.9908 0.8991 0.9687

Ad j ust e d se
s a son a l fa ct ors S t 1 .1 3 4 8 0 .9 9 3 1 0 .9 0 1 1 0 .9 7 1 0

e) Calculate the linear trend from the deseasonalized sales. (9 points)

We apply linear regression with Time Period (t) as independent variable and
the Deseasonlaised Time Series (T, trend line) as the dependent variable.

D e sea son a le d Tim e

Tim e Ser ies Per iod ( 3) = ( 2) -( 2) ' ( 4) = ( 1) - ( 1) ' (3) x (4) ( 3) 2
(1) ( 2)
264.357 1 -7.50 -100.61 754.55 56.25
241.677 2 -6.50 -123.29 801.36 42.25
266.328 3 -5.50 -98.63 542.49 30.25
298.672 4 -4.50 -66.29 298.31 20.25
308.416 5 -3.50 -56.55 197.91 12.25
302.096 6 -2.50 -62.87 157.17 6.25
310.716 7 -1.50 -54.25 81.37 2.25
329.570 8 -0.50 -35.39 17.70 0.25
361.288 9 0.50 -3.68 -1.84 0.25
402.794 10 1.50 37.83 56.75 2.25
432.783 11 2.50 67.82 169.55 6.25
422.261 12 3.50 57.30 200.54 12.25
431.783 13 4.50 66.82 300.69 20.25
453.144 14 5.50 88.18 484.99 30.25
488.268 15 6.50 123.31 801.49 42.25
525.252 16 7.50 160.29 1202.17 56.25
364.9628 8.50 Sum= 6065.19 340.00
(1)' (2)'
b1 = 17.839
b0 = 213.333

The trend line equation is

Tt = 213.333+17.839t

Page 12 of 13
 Split exam: Applied Statistics
Lecturer: Prof. Visieu Lac

f) Forecast the number of units sold in each quarter of year 5. (6

points)

N u m be r of Un it s Tim e Pr e dict ion w it h

Ye a r Qua r t e r Tr e nd lin e
( Y) Pe r iod Se a sona l fa ct or s

1 300 1 231.17 262.34

2 240 2 249.01 247.28
1
3 240 3 266.85 240.47
4 290 4 284.69 276.42
1 350 5 302.53 343.32
2 300 6 320.37 318.14
2
3 280 7 338.20 304.77
4 320 8 356.04 345.71
1 410 9 373.88 424.29
2 400 10 391.72 389.00
3
3 390 11 409.56 369.07
4 410 12 427.40 414.99
1 490 13 445.24 505.27
2 450 14 463.08 459.86
4
3 440 15 480.91 433.37
4 510 16 498.75 484.27
17 516.59 5 8 6 .2 4 516.59 * 1.1348
18 534.43 5 3 0 .7 2
19 552.27 4 9 7 .6 7
20 570.11 5 5 3 .5 5 570.11 * 0.9710

[--------------------------------- End of Paper -----------------------------]

[-------------------------------- End of Paper -------------------------------]

Page 13 of 13