1.

Construct a 98% confidence interval for the average sale growth of all firms
in Asian nations.
Descriptive statistics

salesgrowt
h
count 1318
mean 0,102763
sample standard deviation 0,201465
sample variance 0,040588

confidence interval 98,%

lower 0,089854
confidence interval 98,%
upper 0,115673
margin of error 0,012910
z 2,326
2. Construct a 82% confidence interval for the proportion of manufacturing
firms of all firms in Asian nations
Confidence interval - proportion

82% confidence level

0,463581184 proportion
1318 n
1,341 z
0,018 margin of error
0,445 lower confidence limit
0,482 upper confidence limit

611
x=611, n=1318 => p= =0.4635
1318
Za/2= invNorm(0.91)= 1.341
Therefore, the confidence interval here is given as:

√
p ± Za/2*
p(1− p)
n
= 0.4635 ± 1.341*
√
0.4635(1−4635)
1318
= (0.4451 ; 0.4819)

3. A manager claims that the growth of advertising for Asian firms cannot be
below 10.5%, citing a recent significant global upswing in advertising
expenditure. Is the manager's statement correct at a 10% significance level?
Additionally, what is the percentage of Type I error that would lead us to
reject the null hypothesis?
H0: µ ≤ 10.5%
H1: µ > 10.5%

x = 0.1152, s = 0.4917, n=1318

z=
√ n(x−µ 0) = √ 1318(0.1152−0.1050) =0.7531
s 0.4917
p-value= 0.2256 > α=0.1
 Do not reject
The percentage of Type I error that would lead us to reject the null hypothesis is 10%

4. Assuming equal population standard deviations, a researcher claims that

the salesgrowth of firms in Vietnam is greater than that in Thailand. At a
significance level of α = 0.05, can the researcher's claim be rejected?
Descriptive statistics

Vietnam
count 246
mean 0,101048
sample standard deviation 0,181725

Descriptive statistics

Thailand
count 291
mean 0,095508
sample standard deviation 0,194986

µ1: Salesgrowth of firms in Vietnam, µ2: Salesgrowth of firms in Thailand

H0: µ1 ≤ µ2
H1: µ1 > µ2

THAILAND VIETNAM
n 291 246
x 0.096 0.101
sd 0.195 0.182

√ √
2 2 2 2
s= ( n1−1 )∗s 1 + ( n 2−1 )∗s 2 = ( 246−1 )∗0.1 82 + ( 2 91−1 )∗0.1 9 5 = 0.189
n 1+n 2−2 291+246−2
x 1−x 2 0.101−0. 096
t=
s
√ 1
+
1 =
n1 n2
0.189
1
+
√
1 = -0.3 => p-value = 0.3821 > α=0.05 => fail to reject
291 246
Hence, there is sufficient evidence that salesgrowth of firms in Vietnam is greater than that
in Thailand.

5. At a significance level of α = 0.01, can the researcher's hypothesis that the

advertising expenditures of firms in the retail industry are equal to those in
other industries be rejected, assuming equal population standard deviations?
 Descriptive
statistics

Non-retail
adver
count 659
mean 0,129004
sample standard deviation 0,539565
sample variance 0,291130

Descriptive
statistics

retail adver
count 659
mean 0,101443
sample standard deviation 0,438619
sample variance 0,192386

µ1: Advertising expenditures of firms in retail, µ2: Advertising expenditures of firms in non-
retail
H0: µ1 = µ2
H1: µ1 ≠ µ2

Non-retail Retail
n 659 659
x 0.129 0.101
sd 0.540 0.439

√ √
2 2 2 2
s= ( n1−1 )∗s 1 + ( n 2−1 )∗s 2 = ( 659−1 )∗0.540 + ( 659−1 )∗0.439 = 0.492
n 1+n 2−2 659+659−2
x 1−x 2 0.1 01−1.129
t=
s
√
1
+
n1 n2
1 =
0.492
1
+
√
1 = -1.03
659 659
tα/2 = 2.576
|t|< tα/2 => fail to reject
Hence, there is sufficient evidence that the advertising expenditures of firms in the retail
industry are equal to those in other industries.

6. Determine the main independent variable, the dependent variable, and

control variables.
Dependent variable: salesgrowth
Main independent variable: adver
Control variables: fix, assets_growth, manufacturing, retail, inflation and gdpgrowth.

7. Using the “ga2” dataset to run regression model to test this relationship.
Write the OLS regression function with the estimates for the parameters and
the standard errors under them.
Regression Analysis

R² 0,238
Adjusted
R² 0,234
R 0,488
Std. Error 0,176
n 1318
k 7
Dep. Var. salesgrowth

ANOVA table
Source SS df MS F p-value
Regression 12,7235 7 1,8176 58,46 4,37E-73
Residual 40,7312 1310 0,0311
Total 53,4547 1317

Regression output confidence interval

coefficient t 99% 99%
variables s std. error (df=1310) p-value lower upper
Intercept -0,2198
gdpgrowth 1,1006 0,2018 5,454 5,88E-08 0,5800 1,6211
inflation 0,0697 0,2510 0,278 ,7812 -0,5776 0,7171
fix -0,0101 0,0328 -0,309 ,7574 -0,0946 0,0744
adver 0,0482 0,0101 4,765 2,10E-06 0,0221 0,0743
asset_growth 0,2584 0,0159 16,282 2,00E-54 0,2174 0,2993
manufacturing -0,0102 0,0099 -1,025 ,3056 -0,0358 0,0154
retail 0,0242 0,0099 2,452 ,0143 -0,0013 0,0497

salesgrowth = -0.2086 + 0.0477adver + 1.1028gdpgrowth + 0.0876inflation – 0.0190fix + 0.2564assetgrowth - 0.0102manufacturing +

0.0242retail
(0.0101) (0.2018) (0.2510) (0.0328) (0.0159) (0.0099)
(0,0099)

8. Interpret all coefficients including test significance and economic

magnitude.
The coefficient on adver is 0.0483, positive and statistically significant at 1%. A 1% increase in
advertisement growth is associated with an increase in salesgrowth by 0.0483%.
The coefficient on gdpgrowth is 1.0881, positive and statistically significant at 1%. A 1%
increase in GDP growth is associated with an increase in salesgrowth by 1.0881%.
The coefficient on inflation is 0.2564, positive and statistically significant at 1%. A 1% increase
in asset growth is associated with an increase in salesgrowth by 0.2564%.
The coefficient on fix is -0.0190, negative and statistically significant at 1%. A 1% increase in
fix is associated with an decrease in salesgrowth by 0.019%.
The coefficient on assets_growth is 0.2583, positive and statistically significant at 1%. A 1%
increase in assets growth is associated with an increase in salesgrowth by 0.2583%.

9. Based on your regression from model above, what percentage of the

variation in salesgrowth is explained by independent variables?
R2=0.238 => all independent variables can explain 23.8% of total variations in salesgrowth

10. Calculate the 99% confidence interval for the main independent variable.
Regression output confidence interval
t 99% 99%
variables coefficients std. error (df=1316) p-value lower upper
Intercept 0,0932
adver 0,0830 0,0111 7,501 1,16E-13 0,0544 0,1115

11. Check whether multicollinearity exists in your model.

Regression output confidence interval
coefficient t 95% 95%
variables s std. error (df=1310) p-value lower upper VIF
Intercept -0,2198
gdpgrowth 1,1006 0,2018 5,454 5,88E-08 0,7047 1,4964 1,273
inflation 0,0697 0,2510 0,278 ,7812 -0,4226 0,5620 1,318
fix -0,0101 0,0328 -0,309 ,7574 -0,0744 0,0541 1,094
adver 0,0482 0,0101 4,765 2,10E-06 0,0284 0,0681 1,048
asset_growth 0,2584 0,0159 16,282 2,00E-54 0,2272 0,2895 1,068
manufacturing -0,0102 0,0099 -1,025 ,3056 -0,0296 0,0093 1,038
retail 0,0242 0,0099 2,452 ,0143 0,0048 0,0436 1,035
1,125
mean
VIF
VIF < 2 => no multicollinearity

12. Calculate mean and standard deviation of expected value of salesgrowth

and residuals. Are the results consistent with multiple regression model
assumptions?
Descriptive statistics

Predicted
count 1318
mean 0,1027635
sample standard deviation 0,0982904
sample variance 0,0096610

Descriptive statistics

Residual
 count 1318

mean -0,0000000

sample standard deviation 0,1758614

sample variance 0,0309272

Assumptions:
1. Linear relationship: There exists a linear relationship between each predictor
variable and the response variable.

1.20000
f(x) = x
1.00000 R² = 0.238024746498759
0.80000
0.60000
salesgrowth

0.40000
0.20000
0.00000
-0.50000 0.00000 0.50000 1.00000 1.50000 2.00000 2.50000
-0.20000
-0.40000
-0.60000
Predicted

2. No Multicollinearity: None of the predictor variables are highly correlated

with each other.
VIF<2 => no multicollinearity

3. Independence: The observations are independent.

Durbin-Watson = 2.05
d > 2 indicates negative serial correlation but if d is between 1.5 and 2.5 then autocorrelation
is likely not a cause for concern.

4. Homoscedasticity: The residuals have constant variance at every point in the

linear model.
 8

4
Standard Residuals

0
-0.5 f(x) =0 − 4.13490933390435E-15
0.5 1 x −1.5
1.91716333392304E-17
2 2.5
R²-2= 0

-4

-6

-8
Predicted Y

5. Multivariate Normality: The residuals of the model are normally distributed.

Histogram
40
35
30
25
20
Percent

15
10
5
0
0 0 0 0 0 0 0 0 0 0 0 0
000 000 000 000 000 000 000 000 000 000 000 000
.0 0.0 0.0 0.0 0.0 0.0 9.0 0.0 0.0 0.0 0.0 1.1
-1
Residual

13. Can we reject the hypothesis the coefficient on fix = −0.2 against two-
sided alternative at 5% significance level using three methods?
14. At 1% significance level, test the hypothesis that a 1% increase in GDP
growth leads to 1% increase in firm salesgrowth.
15. To optimize the model, we may consider to drop one or several variables.
Which variables should we drop?
Regression output confidence interval
coefficient t 95% 95%
variables s std. error (df=1310) p-value lower upper VIF
 Intercept -0,2198
gdpgrowth 1,1006 0,2018 5,454 5,88E-08 0,7047 1,4964 1,273
inflation 0,0697 0,2510 0,278 ,7812 -0,4226 0,5620 1,318
fix -0,0101 0,0328 -0,309 ,7574 -0,0744 0,0541 1,094
adver 0,0482 0,0101 4,765 2,10E-06 0,0284 0,0681 1,048
asset_growth 0,2584 0,0159 16,282 2,00E-54 0,2272 0,2895 1,068
manufacturing -0,0102 0,0099 -1,025 ,3056 -0,0296 0,0093 1,038
retail 0,0242 0,0099 2,452 ,0143 0,0048 0,0436 1,035
1,125
mean
VIF
H0: 𝛽 = 0;
H1: 𝛽 ≠ 0;
At 5% significance level => α=0.5
p-value of inflation (0.7812), fix (0.7574) and manufacturing (0.3056) are higher than α.
=> fail to reject H0 at 5% significance level
Therefore, inflation, fix and manufacturing have insignificant impacts on salesgrowth.
=> we can drop these variables

16. Suppose we create a new variable 𝑛𝑜𝑛_𝑟𝑒𝑡𝑎𝑖𝑙 = 1 − 𝑟𝑒𝑡𝑎𝑖𝑙. Can we include

this variable in the above regression? Please explain.
If we create a new variable non_retail = 1 - retail, cannot include it in the above regression.
There are two values 0 and 1 in the retail column which can be considered as a dummy
variable - a variable that equals 0 or 1. Including this variable in the regression analysis can
bring about disadvantages. This is because it can lead to a dummy trap and when using
dummy variables, one category always has to be omitted. Alternatively, one could omit the
intercept but it is more difficult to test for differences between the parameters. Additionally,
the R-squared formula is only valid if regression contains an intercept.

17. Estimate these two following models and report the estimation results in
equation forms:
𝑠𝑎𝑙𝑒𝑠_𝑔𝑟𝑜𝑤𝑡ℎ = 𝛽0 + 𝛽1𝑎𝑑𝑣𝑒𝑟 + 𝛽2𝑔𝑑𝑝_𝑔𝑟𝑜𝑤𝑡ℎ + 𝛽3𝑖𝑛𝑓𝑙𝑎𝑡𝑖𝑜𝑛 + 𝑢; (2)
𝑠𝑎𝑙𝑒𝑠_𝑔𝑟𝑜𝑤𝑡ℎ = 𝛽0 + 𝛽1𝑎𝑑𝑣𝑒𝑟 + 𝛽2𝑓𝑖𝑥 + 𝛽3𝑎𝑠𝑠𝑒𝑡_𝑔𝑟𝑜𝑤𝑡ℎ +
𝛽4𝑚𝑎𝑛𝑢𝑓𝑎𝑐𝑡𝑢𝑟𝑖𝑛𝑔 + 𝛽5𝑟𝑒𝑡𝑎𝑖𝑙 + 𝑢; (3)
Model (2):

Regression Analysis

R² 0,079
Adjusted R² 0,077
R 0,281
Std. Error 0,194
n 1318
k 3
 Dep. Var. salesgrowth

ANOVA table
Source SS df MS F p-value
3,22E-
Regression 4,2098 3 1,4033 37,44 23
Residual 49,2449 1314 0,0375
Total 53,4547 1317

Regression output confidence interval

t 99% 99%
variables coefficients std. error (df=1314) p-value lower upper VIF
Intercept 0,0636
gdpgrowth 1,4000 0,2202 6,358 2,81E-10 0,8320 1,9679 1,257
inflation 0,1096 0,2700 0,406 ,6850 -0,5870 0,8061 1,266
adver 0,0791 0,0109 7,261 6,56E-13 0,0510 0,1073 1,009
1,178
mean
VIF
salesgrowth = 0.0636 +0.0791adver + 1.4gdpgrowth + 0.1096inflation
(0.0109) (0.2202) (0.2700)

Model (3):

Regression Analysis

R² 0,216
Adjusted R² 0,213
R 0,464
Std. Error 0,179
n 1318
k 5
Dep. Var. salesgrowth

ANOVA table
Source SS df MS F p-value
8,09E-
Regression 11,5272 5 2,3054 72,14 67
Residual 41,9275 1312 0,0320
Total 53,4547 1317

Regression output confidence interval

coefficient t 99% 99%
variables s std. error (df=1312) p-value lower upper VIF
Intercept -0,2152
fix 0,0159 0,0326 0,488 ,6254 -0,0682 0,1001 1,055
adver 0,0500 0,0102 4,889 1,14E-06 0,0236 0,0763 1,040
asset_growth 0,2699 0,0160 16,894 4,36E-58 0,2287 0,3111 1,053
manufacturing -0,0110 0,0100 -1,102 ,2705 -0,0368 0,0148 1,025
retail 0,0251 0,0100 2,508 ,0123 -0,0007 0,0509 1,034
1,041
mean
VIF
salesgrowth = -0.2152 + 0.05adver + 0.0159fix + 0.2699asset_growth 0.011manufacturing + 0.0251retail
 (0.0102) (0.0326) (0.0160) (0.0100) (0.0100)

18. Among the models in part (6), model (2), and model (3), which model is
the best to explain the relationship between advertisement and salesgrowth?
Why?
Part (6) model:
salesgrowth = -0.2086 + 0.0477adver + 1.1028gdpgrowth + 0.0876inflation – 0.0190fix + 0.2564assetgrowth - 0.0102manufacturing +
0.0242retail
(0.0101) (0.2018) (0.2510) (0.0328) (0.0159) (0.0099)
(0,0099)

Adjusted R2 = 0.238
Model (2):
salesgrowth = 0.0636 +0.0791adver + 1.4gdpgrowth + 0.1096inflation
(0.0109) (0.2202) (0.2700)

Adjusted R2 = 0.077
Model (3):
salesgrowth = -0.2152 + 0.05adver + 0.0159fix + 0.2699asset_growth 0.011manufacturing + 0.0251retail
(0.0102) (0.0326) (0.0160) (0.0100) (0.0100)

Adjusted R2 = 0.213

Part (6) model has the highest adjusted R^2 => the model that best explain the relationship between
advertisement and salesgrowth.