Md.

Khaleduzzaman Monayam Hossain Moin Serajul Islam

20231164 20231160 20231167

Mathematical Assignment
&
Project
On
Interest Rate Effect on Investment Growth

Sl. Group Member Name Student Number

1 MD. Khaleduzzaman 20231164
2 Serajul Islam 20231167
3 Monayam Hossain Moin 20231160

Subject: Quantitative methods in Finance

Submission Date: 2nd January 2024

Submitted to: Douglas Eduardo Turatti

1|Page
Md. Khaleduzzaman Monayam Hossain Moin Serajul Islam
20231164 20231160 20231167

Part 1 - Mathematical Assignment

Question 1.1
Explain how to find the weights wA and wB that lead to minimum variance. In other words,
explain the procedure for single (or multiple)-variable unconstrained optimization applied to
the variance of the portfolio. State first-order and second-order conditions applied to this
problem. Note: You do not need to solve first-order and second order conditions.
Answer 1.1:
For this problem we want to find the minimum variance portfolio which is the combination of
the two assets with the lowest risk (variance). This does not work if the two assets have a
perfect positive correlation of 1 because then you won't get the benefit of diversification,
unless you are short-selling the asset with the higher variance which can be hard in practice.
If there is a perfect negative correlation of -1 we can find a combination of asset A and asset
B where the only risk is the systematic market risk which can’t be diversified away. Finding
two stocks with a correlation of -1 is almost impossible in practice. Examples of systematic
market risk could be high interest rates, high unemployment or high inflation which are
external factors that companies can’t directly affect and which does not only affect the given
companies but the whole market.

To find the weights WA and WB that lead to the minimum variance portfolio, we can use
single-variable unconstrained optimization. This means finding the values of the variables
(WA and WB) that minimize or maximize the objective function (in this case, minimizing the
variance of the portfolio).

The procedure for single-variable unconstrained optimization follows these steps:

- Identification of the objective function that we want to minimize or maximize. In this
case, the objective function is the variance of the portfolio, which is a function of WA
and WB.
- Rewriting the objective function as a function of only one variable. In this case, we

2|Page
Md. Khaleduzzaman Monayam Hossain Moin Serajul Islam
20231164 20231160 20231167

can rewrite the variance of the portfolio as a function of only WB.

- Taking the derivative of the objective function with respect to the variable of interest
(WB). This will give us the slope of the objective function.
- Setting the derivative equal to zero and solving for the variable of interest (WB). This
will give us the value of the variable that minimizes the objective function. This results
in the equation:

- Putting the value of the variable back into the objective function to find the minimum
value of the objective function. So, after solving this equation it is easy to find the
weight of asset B because we know that WA + WB = 1. Thus, the weight of asset B is
given by (1-WA ).

Question 1.2
The exponentially weighted moving-average (EWMA) is a popular and simple model to
generate forecasts of covariance matrices. The model is usually written as
ˆΣ
t+1 = (1 − λ)r′
trt + λˆΣt, (3)
where ˆΣt+1 is the forecast for the covariance matrix at day t + 1, λ is a discounting
parameter, rt is a row vector of log-returns for several assets for a given day t, and ˆΣt is the
forecast of the covariance matrix at day t. Provide the forecast of the covariance matrix for
02-11-2023 using the above data set. Consider λ = 0.92, and ˆΣt as the estimated covariance
matrix using the whole data.
Hint: Use a software. Check slides of lecture 3 to _nd how to estimate the covariance matrix.
Identify what is rt, do the matrix multiplication, and _nd the forecast as a sum of two parts.
Answer 1.2:
We have done this forecasting of the covariance matrix for 02-11-2023 in excel spreadsheet
and organize the log-return data for asset A and asset B.
First we calculate rt′rt by multiplying the transpose of the row vector rt by itself. Use the
following formula in a new column for each day
3|Page
Md. Khaleduzzaman Monayam Hossain Moin Serajul Islam
20231164 20231160 20231167

=MMULT(TRANSPOSE(log_returns_t), log_returns_t)
Then we calculated the estimated covariance matrix Σ^t in separate cells. We also input
lamba value 0.92 in a separate cell.
After that we calculate the two parts of the EWMA formula in Question 2, Answer
spreadsheet in F column and G-I Column.
Finally we sum the two parts to get the forecasted covariance matrix for 02-11-2023 Σ^t+1
Which is

Forecast of the covariance matrix

for 02-11-2023 Σ^t+1
-
- 0.9795
0.294636288 -0.3463 2
- -
0.0164 0.2946
-0.34629916 8 4
- -
- 0.3171 0.2218
0.979519156 6 6

Question 1.3
Find the weights wA and wB that lead to minimum variance portfolio for 02-11-2023 giving
your forecast of the covariance matrix in 1.2 and your answer in 1.1.
Show your calculation in details.
Answer 1.3:
Given Values:
Mean A=−0.011201831
Mean B=−0.0031646
Variance A=0.300999519
Variance B=0.082197492
Covariance AB=−0.024107041
The portfolio variance is given by:
2 2 2 2 2
σ p=w A σ A +w B σ B+2 wAwBσAB
2 2 2
σ p=w Variance A +w BVariance B+2 wAwBCocariance AB
Substitute Values:

4|Page
Md. Khaleduzzaman Monayam Hossain Moin Serajul Islam
20231164 20231160 20231167
2 2 2
σ p=w ·0.300999519+ w B· 0.082197492+2 wAwB· 0.024107041
Substitute wA in Terms of wB:
Using the constraint wA+wB=1, substitute wA=1−wB into the portfolio variance formula.
Simplify the Objective Function:
Combine like terms and simplify the expression.
Take the Derivative:
Take the derivative of the simplified objective function with respect to wB.
Set the Derivative to Zero:
Set the derivative equal to zero and solve for wB.
Check Second-Order Condition:
Take the second derivative and check its sign.
2 2 2
σ p=w Variance A +w BVariance B+2 wAwBCocariance AB
2 2 2
σ p=w ·0.300999519+ w B· 0.082197492+2 wAwB· 0.024107041
2 2
σ p=0.300999519−.600314338 wB+0.281512533 w B−0.048214083 wB
Combine like terms:
2 2
σ p=0.281512533 w B−0.648528421 wB +0.300999519
Take the Derivative:
2
dσ p
=0.563025066 wB−0.648528421=0
dwB
Solve for wB:
0.648528421
wB = ≈ 1,151864
0.563025066
Check Second-Order Condition:
The second derivative is positive, indicating a minimum.
Solve for wA:
Using the constraint wA+wB=1:
wA=1−wB
wA=1−1,151864=−0,151864

Question 2.1 Write the matrix X in the context of model (5). What is the dimension of this
matrix?

5|Page
Md. Khaleduzzaman Monayam Hossain Moin Serajul Islam
20231164 20231160 20231167

Answer 2.1:
The matrix X is constructed by adding a column of ones for the intercept term (β0) and
appending the columns for x1 and x2.

1 x1 x2
0.91476
1 3 1.85633
- 0.02915
1 0.79394 2
-
1 0.99536 -1.5624
- -
1 1.56614 0.52496
0.02453 0.29048
X= 1 2 3
- -
1 0.44527 0.17613
- -
1 0.69433 0.47672
0.83741 -
1 7 0.88305
0.18905 -
1 5 0.34813
1.26889 -
1 2 0.08649

The dimension of this matrix is 10×3 (10 observations, 3 columns).

Question 2.2 What is the dimension of the matrix y.

Answer 2.2:
The matrix y is the column vector of the regressand (response variable) y.
y
12.315
6
0.2111
12
-
6.4599

6|Page
Md. Khaleduzzaman Monayam Hossain Moin Serajul Islam
20231164 20231160 20231167

3
-
5.0539
6
4.0159
37
0.4382
56
-
1.5706
5
1.0984
72
1.7673
14
5.7966
91

The dimension of this matrix is 10×1 (10 observations, 1 column).

Question 2.3 Find and write the matrix (X′X)−1. What is the dimension of this matrix?
Hint: In R, matrix inversion can be easily performed using inv() from the package matlib.
Answer 2.3:
To find (X′X)−1, we need to perform matrix multiplication in Excel. Which I have shown in
excel spradesheet.
Transpose of X (X′):
Create a new matrix with the transpose of X.
12.3156 0.211112 -6.45993 -5.05396 4.015937 0.438256 -1.57065 1.098472 1.767314 5.796691
0.914763 -0.79394 -0.99536 -1.56614 0.024532 -0.44527 -0.69433 0.837417 0.189055 1.268892
1.85633 0.029152 -1.5624 -0.52496 0.290483 -0.17613 -0.47672 -0.88305 -0.34813 -0.08649

Matrix Multiplication (X′X):

Multiply the transposed X by X.
275.710 35.3655
1 35.0467 5
3.55375
35.0467 7.93872 8
35.3655 3.55375 7.41459
5 8 2

7|Page
Md. Khaleduzzaman Monayam Hossain Moin Serajul Islam
20231164 20231160 20231167

Matrix Inversion ((X′X)^-1:

Invert the resulting matrix.
0.01834 - -
6 0.05324 0.06199
- 0.31489 0.10302
0.05324 4 8
- 0.10302 0.38114
0.06199 8 5

The dimension of (X′X)−1 is 3×3.

Question 2.4 Find and write the matrix X′y. What is the dimension of this matrix?
Answer 2.4:
Here, we multiply the X matrix with the y matrix.
275.710
1
35.0467
35.3655
5

The matrix multiplication can be done because the number of columns in matrix X is equal
to the number of rows in matrix y (10), and the resulting matrix has the number of rows of X,
(3) and the number of columns of y (1).
This matrix is a column vector with the dimension (3x1)

Question 2.5 Now multiply (X′X)−1 and X′y to find ˆ β. What is the dimension of ˆ β?
Explain.
Answer 2.5:
Perform the matrix multiplication:
1.0925
-1.2075
0.5175

8|Page
Md. Khaleduzzaman Monayam Hossain Moin Serajul Islam
20231164 20231160 20231167

The dimension of β is 3×1. This is because it consists of the estimates for β, β1, and β2.

Question 2.6 Compare your calculation in 2.5 with a built-in command such as lm() in R or
the respective in Excel. Are they exactly the same?
Answer 2.6:
The Excel command Linest() is used to find βˆ.

The results are exactly the same, which confirms the results found by using the OLS
formula.

Part 2-Empirical Study

Introduction:
The relationship between USA interest rates and investment growth is a topic of ongoing
debate and research within the fields of economics and finance. At first glance, it may seem
like these two variables are unrelated, but in fact, they are deeply interconnected and can

9|Page
Md. Khaleduzzaman Monayam Hossain Moin Serajul Islam
20231164 20231160 20231167

have significant effects on each other. For investors, understanding this relationship is
essential for making rational and informed decisions about allocation of their capital. For
policymakers, the relationships are critical for the designing of monetary and fiscal policies
that can effectively stimulate economic growth and maintain price stability. Central banks
worldwide use the flexibility of interest rates as a tool to manage inflation and the economy if
there are signs of an economic slowdown and rising unemployment. If the economy is
overheating, the central banks can increase the interest rate to cool down the economy in
hopes of avoiding recession. When these economic factors change, the companies worldwide
are affected both directly and indirectly. The performance of public companies are reflected
in their stock price, which are usually affected alongside with the company’s overall
performance. The thesis that forms the basis for this paper is that large companies perform
better than mid-sized and small companies during periods of elevated interest rates and
inflation. Our overall belief is that large companies have solid relationships with their
customers and have the advantage of economies of scale. Smaller companies do not have
decade-long relationships with customers and their businesses might not take up that much
space in the customers’ minds. Furthermore, they cannot just raise their prices without the
risk of losing a significant number of customers. Large companies have easier access to the
capital markets too, given their large amount of assets and their long relationships with banks.
In this empirical analysis paper, we will delve into the intricate and often dynamic
relationship between stock returns, inflation, and interest rates. We will analyze the
performance of different sized companies from different parts of the world during periods of
changing interest rates and inflation to test our thesis that the stock returns of large companies
are less negatively affected by increasing interest rates and inflation than smaller companies.

Nowadays in globalized world relationship between interest rates and investments has a great
demand to financialist, economist, managers etc. In this globalized world USA is the top
country who plays a key role to shape this economy. In 2013-2022 interest rates in United
States of America had several fluctuations. They done this to reconstruct investment
strategizes in every sectors. In this empirical study we focused on the connection between
United States of America’s interest rates and investment growth in oil, telecommunication,
and food industries. All of these industries are different in their field and their investment
decision is also different depends on interest rate. The oil industry, which is very important to
the world's energy markets, is volatile due to a combination of geopolitical and economic
concerns. On the other hand, the telecommunications sector is an very important part of the
contemporary information era and is subject to changes in regulations and technical
improvements. In the meanwhile, shifting consumer tastes and global supply chain issues
affect the food business, which is a vital sector for human nourishment.
From 2013-2022, in these ten years of period we will conduct correlations and causal
relationships between interest rates up and down and investment growth and decrease in these
three industries. This analysis will help us to determine how cost of capital changes effect in
decision making, capital expenditure etc in these oil, telecommunication and food industry.
Our goal as we set out on this empirical study is to add significant knowledge to the
conversation about how interest rates and industry investments behave each other. In addition
to advancing our knowledge of economic dynamics, our research offers useful advice for

10 | P a g e
Md. Khaleduzzaman Monayam Hossain Moin Serajul Islam
20231164 20231160 20231167

companies, investors, and decision-makers negotiating the complex web of financial choices
in a quickly changing economic environment.
We will start by presenting the data we have gathered, which will be used throughout the
article, along with the techniques we employed to process the information and develop our
thesis. This will give the reader and us a strong basis on which to build our knowledge of the
ways in which interest rates and investment behavior interact. Next, we will conduct the
analysis of our data, examining how investment and interest rates have interacted in the past
ten years for our selected companies in oil, telecommunication and food industry. We will
also be able to say something about how they may be expected to behave in the future. By the
end of this paper, we aim to have provided a thorough and nuanced understanding of the
relationship between investment and interest rates. Through our analysis, we hope to shed
light on the complex and multifaceted nature of this relationship and to provide insights that
may be useful for investors and policymakers.
Scope and Limitations:
Since our project on USA interest and investment growth of 15 companies in three industries
we still have more scope to find further information. But because of the time limitation,
information, accessibility, scarcity of data availability we face difficulties to complete the
project on best way.

Methodology:

Data Construction:
Our study focuses on three major industries—Oil, Telecommunication, and Food—spanning
the years 2013 to 2022. For each industry, we have selected five key companies, providing a
representative sample for analysis. The companies include Chevron, Exxon Mobil,
ConocoPhillips, Marathon Petroleum, and Valero Energy in the Oil sector; Verizon, T-Mobile,
Comcast Corp, AT&T, and Charter Communications in the Telecommunication sector; and
Campbell Soup Company, General Mills, Mondelez International, PepsiCo, and Conagra in
the Food sector.
Dataset Description:
To facilitate our analysis, we constructed panel datasets in long format, organizing
observations chronologically across time. Each dataset corresponds to one industry Oil,
Telecommunication, or Food. These datasets contain information on total investments
(dependent variable) and U.S. interest rates (independent variable) for each company,
creating a comprehensive framework for our analysis.
Independent Variable in Dataset:

Year Interest Rate (%)

2013 1.5

11 | P a g e
Md. Khaleduzzaman Monayam Hossain Moin Serajul Islam
20231164 20231160 20231167

2014 1.4
2015 2.2
2016 2.5
2017 2.2
2018 2.4
2019 3.4
2020 2.2
2021 -1.2
2022 1.58

Dependent Variable (Growth) in Dataset:

Oil Industry:

Chevron Exxon Mobil

Total Total
Yea investments Growt Yea investments Growt
r (millions of h r (millions of h
dollars) dollars)
2013 25,502 2013 36,328
2014 26,912 6% 2014 35,239 -3%
2015 27,110 1% 2015 34,245 -3%
2016 30,250 12% 2016 35,102 3%
2017 32,497 7% 2017 39,160 12%
2018 35,546 9% 2018 40,790 4%
2019 38,688 9% 2019 43,164 6%
2020 39,052 1% 2020 43,515 1%
2021 40,696 4% 2021 45,195 4%
2022 45,238 11% 2022 34,718 -23%

ConocoPhillips Marathon Petroleum

Year Total Growt Year Total Growth
investments h investments
(millions of (millions of

12 | P a g e
Md. Khaleduzzaman Monayam Hossain Moin Serajul Islam
20231164 20231160 20231167

dollars) dollars)
2013 23,907 2013 1,800
2014 24,335 2% 2014 1,744 -3%
2015 20,490 -16% 2015 1,570 -10%
2016 21,091 3% 2016 1,659 6%
2017 9,599 -54% 2017 1,840 11%
2018 9,329 -3% 2018 2,089 14%
2019 8,687 -7% 2019 2,531 21%
2020 8,017 -8% 2020 2,621 4%
2021 7,113 -11% 2021 3,043 16%
2022 8,225 16% 2022 1,838 -40%

Valero Energy
Total investments
Yea
(millions of Growth
r
dollars)
2013 4,292
2014 3,689 -14%
2015 4,114 12%
2016 4,816 17%
2017 5,850 21%
2018 4,850 -17%
2019 2,938 -39%
2020 2,490 -15%
2021 2,467 -1%
2022 2,738 11%

Telecommunication Industry:

Varizon T Mobile
Total Total
Yea investments Growt Yea investments Growt
r (millions of h r (millions of h
dollars) dollars)
201 201
3 16,604 3 4,025

13 | P a g e
Md. Khaleduzzaman Monayam Hossain Moin Serajul Islam
20231164 20231160 20231167

201 201
4 17,191 4% 4 4,317 7%
201 201
5 17,775 3% 5 4,724 9%
201 201
6 17,059 -4% 6 4,702 0%
201 201
7 17,247 1% 7 5,237 11%
201 201
8 16,658 -3% 8 5,541 6%
201 201
9 17,939 8% 9 6,391 15%
202 202
0 18,192 1% 0 11,034 73%
202 202
1 20,286 12% 1 12,326 12%
202 202
2 23,087 14% 2 13,970 13%

Comcast Crop AT & T

Total Total
Yea investments Growt Yea investments Growt
r (millions of h r (millions of h
dollars) dollars)
2013 5,403 2013 21,228
2014 6,154 14% 2014 21,433 1%
2015 7,034 14% 2015 20,015 -7%
2016 7,596 8% 2016 22,408 12%
2017 7,952 5% 2017 21,550 -4%
2018 7,716 -3% 2018 21,251 -1%
2019 6,909 -10% 2019 19,635 -8%
2020 6,605 -4% 2020 15,675 -20%
2021 6,930 5% 2021 16,527 5%
2022 7,568 9% 2022 19,626 19%

14 | P a g e
Md. Khaleduzzaman Monayam Hossain Moin Serajul Islam
20231164 20231160 20231167

Charter Communications
Total
Yea investments Growt
r (millions of h
dollars)
2013 1,854
2014 2,221 20%
2015 1,840 -17%
2016 7,545 310%
2017 8,681 15%
2018 9,125 5%
2019 7,195 -21%
2020 7,415 3%
2021 7,635 3%
2022 9,376 23%

Food Industry:

Campbell Soup Company General Mills

Total Total
Yea investments Growt Yea investments
Growth
r (millions of h r (millions of
dollars) dollars)
2013 761 2013 -40.40
2014 821 8% 2014 -54.90 36%
2015 805 -2% 2015 -102.40 87%
2016 691 -14% 2016 63.90 -162%
2017 666 -4% 2017 3.30 -95%
2018 631 -5% 2018 -17.30 -624%
2019 644 2% 2019 0.10 -101%
-
2020 568 -12% 2020 -48.00 48100%
2021 702 24% 2021 -76.40 59%
2022 534 -24% 2022 14.70 -119%

15 | P a g e
Md. Khaleduzzaman Monayam Hossain Moin Serajul Islam
20231164 20231160 20231167

Mondelez International Pepsico

Total Total
Yea investments Growt Yea investments Growt
r (millions of h r (millions of h
dollars) dollars)
2013 4,466 2013 104
2014 3,932 -12% 2014 -17 -116%
2015 8,902 126% 2015 83 -588%
2016 9,484 7% 2016 -39 -147%
2017 10,973 16% 2017 157 -503%
2018 9,908 -10% 2018 -77 -149%
2019 11,430 15% 2019 -30 -61%
2020 12,802 12% 2020 235 -883%
2021 12,724 -1% 2021 -192 -182%
2022 8,538 -33% 2022 -120 -38%

Conagra
Total
Yea investments Growt
r (millions of h
dollars)
2013 15.20
2014 12.60 -17%
2015 853.00 6670%
2016 910.80 7%
2017 1,305.00 43%
2018 700.00 -46%
2019 245.90 -65%
2020 241.00 -2%
2021 263.40 9%
2022 230.80 -12%

Regression Analysis:
Our analytical approach involves single regression analysis for each industry. We are doing
this single regression analysis with industry average. We adopt a single regression model with
investment growth of U.S three industries’ as the dependent variable and U.S. interest rates as
the independent variable. The model is expressed as follows:
y = mx + c,
where

16 | P a g e
Md. Khaleduzzaman Monayam Hossain Moin Serajul Islam
20231164 20231160 20231167

m indicates slope
c indicates y-intercept
x indicates independent variable
y indicates dependent variable

We assume this equation to be the Data Generating Process (DGP) for y, and that y is created
by this mechanism in the real world (Wooldridge, 2015). This will provide us with equations
for each tier of industries and give us an idea of how they are performing under certain levels
of interest rates. Our independent variables are the interest rates (x) and our dependent
variables are the average investment growth of industries (y). We want to estimate the slope
perimeters to analyze the change in y given a unit change in x.
Diagnostic Tests:
We conduct following three crucial diagnostic tests to validate our regression results:
1. Jarque Bera Test:
A test for normalcy that is a variation of the Lagrange multiplier test is the Jarque-Bera
test. Many statistical tests, such as the t test and the F test, rely on the assumption of
normalcy. To verify normalcy, the Jarque-Bera test is typically conducted before to one of
these tests. Since alternative normality tests become unreliable for big n values (Shapiro-
Wilk, for instance, becomes unreliable for n greater than 2,000), it is typically used to
huge data sets.
To be more precise, the test compares the data's skewness and kurtosis to check if it
fits a normal distribution. The data may be in any of the following formats:
Time Series Data.
Errors in a regression model.
Data in a Vector.
A normal distribution has a kurtosis of three, which indicates how much of the data is
in the tails and how "peaked" the distribution is, and a skew of zero, which indicates
that it is entirely symmetrical around the mean. To conduct the test, one does not need
to know the data's mean or standard deviation.
2. Heteroskedasticity Test:
In statistics, the term heteroscedasticity is the opposite of homoscedasticity which is used to
characterize situations in which the variance of the model's errors varies for each observation.
This is because a fundamental tenet of modeling is frequently that the variances and the
model's errors have the same distribution. This is especially true when it comes to linear
regression or time series analysis.

17 | P a g e
Md. Khaleduzzaman Monayam Hossain Moin Serajul Islam
20231164 20231160 20231167

Because the model's errors, also known as residuals, in linear regression analysis are not
homoskedastic, the model coefficients estimated using ordinary least squares (OLS) are
neither unbiased nor those with the least variance. Their variance cannot be reliably
estimated.

 It is thus required to do a test for heteroscedasticity if it is believed that the variances

are not homogenous (a representation of the residuals versus the explanatory variables
may indicate heteroscedasticity). The null and alternative hypotheses for the
following tests have been developed:

White test and modified White test

How can the White test be used to identify heteroscedasticity? White (1980) created this
technique to find instances of heteroscedasticity that undermine the accuracy of traditional
estimators of linear regression parameters. Although the concept is similar to Breusch and
Pagan's, it is predicated on less robust hypotheses on the shape of heteroscedasticity. As a
consequence, the explanatory factors, as well as the squares and cross-products of those
variables, are used to regression the quadratic errors.

Goodness of Fit Test:

There are several reasons why goodness-of-fit tests are significant in statistics. This test aids
in evaluating how well a statical model fits a given collection of empirical data. This
goodness of fit test's primary goal is to determine whether or not the observed data are
compatible with the statical model that is being assumed. A good model may be chosen from
a variety of models by using the goodness fit test.

Finding market abnormalities or outliers that may be affecting how well the model fits the
data may also be done with this test. Outliers may need to be removed or treated differently
since they might have a substantial impact on the model fit. Sometimes outliers are hard to
spot unless they are part of an analytical model.

Tests for goodness-of-fitness might also reveal details on the data's variability and the
model's predicted parameters. Predictions and a knowledge of the behavior of the system
being modeled can both benefit from this information. The model unique to the dataset
being tested, the residuals being computed, and the p-value for possibly extreme data may
need to be refined based on the data being fed into the model.
By implementing this methodology, we aim to provide a comprehensive understanding of
how U.S. interest rates influence investment growth in the Oil, Telecommunication, and Food
industries across different company sizes from 2013 to 2022.

18 | P a g e
Md. Khaleduzzaman Monayam Hossain Moin Serajul Islam
20231164 20231160 20231167

Analysis of empirical results

Assumptions:

1. Linear in parameters
Since this is a linear regression model, the coefficients are multiplied by the independent
variables, X1 (interest rates), to determine whether there is a linear relationship between Y
(investment growth) and the independent variables. Scatter plot diagrams are the most
straightforward method for determining whether or not this assumption holds true for our
model. This makes it possible for us to see whether or not the assumption is met visually.
There may be a linear relationship if the points of the X's and Y's are near to a straight line.
There may not be a linear relationship if the points are dispersed and do not form a relative
straight line. Finding a straight line that minimizes the sum of the residuals is the aim of
linear regression.
Scatterplots of the following

Residuals
12

10

8 Residuals
Linear (Residuals)
6

0
0 2 4 6 8 10 12

19 | P a g e
Md. Khaleduzzaman Monayam Hossain Moin Serajul Islam
20231164 20231160 20231167

Residuals
12

10

8
Residuals
Linear (Residuals)
6

0
0 2 4 6 8 10 12

Residuals
12

10

8
Residuals
Linear (Residuals)
6

0
0 2 4 6 8 10 12

2. Correct specification
Our model links our independent variables X (interest rates) to the dependent variable y
(invest growth). No other variable possesses the ability to consistently impact our outcome.
Every variable in our model serves as a pertinent explanatory variable. In other words, the
model has no redundant, superfluous, or missing predictors. Model misspecification occurs
when this is done incorrectly, with coefficients being assigned to variables that do not exist in
the model's data set, leading to incorrect results.

20 | P a g e
Md. Khaleduzzaman Monayam Hossain Moin Serajul Islam
20231164 20231160 20231167

The way that interest rates and investment growth are related worries us. Our slope with
respect to our coefficients is neither over- nor under-specified for any of the three models
because these are interconnected.

Oil Industry

Regression Statistics
Multiple R 0.114476666
R Square 0.013104907
Adjusted R Square -0.127880106
Standard Error 0.042379532
Observations 9

1. Multiple R:
 The Multiple R which is the correlation coefficient between the actual values
of the dependent variable and the predicted values. The value is nearly 0.1145.
It presents the solidity and path of the linear correlation between the
independent and dependent variables.
2. R Square (R²):
 The value of R Square is the coefficient of determination. R² is roughly
0.0131, indicating that only about 1.31% of the variability in the dependent
variable can be describe by the independent variables in the model.
3. Adjusted R Square:
 The Adjusted R Square is a revised model of R² that adjusts for the number of
independent variables in the model. It penalizes the inclusion of irrelevant
variables that do not contribute much to the interpretation of the dependent
variable. In this case, the Adjusted R Square is approximately -0.1279, which
suggests that the model is not so well-adapted to the data.
4. Standard Error:
 The Standard Error is a evaluation of the average distance between the stated
values as well as the values forecasted by the model. In our case, it is around
0.0424.
5. Observations:
 This shows the number of data points or observations used in the regression
analysis. In this model, there are 9 observations.
It's important that an Adjusted R Square value below zero is uncommon and shows minor
issues with the model, such as overfitting or containing some irrelevant variables.

21 | P a g e
Md. Khaleduzzaman Monayam Hossain Moin Serajul Islam
20231164 20231160 20231167

Additionally, the low R Square value suggests that the model does not explain whole of the
variability in the dependent variable.
ANOVA
df SS MS F Significance F
Regression 1 0.000166945 0.000166945 0.092952483 0.769321884
Residual 7 0.012572173 0.001796025
Total 8 0.012739118

1. Regression:
 df (Degrees of Freedom): df equals to 1, indicating the number of independent
variables in the model which is 1 in simple linear regression.
 SS (Sum of Squares): The sum of squared distinct between the forecasted
values and the mean of the dependent variable. In this case, SS =
0.000166945.
 MS (Mean Square): The mean square, calculated as SS divided by the degrees
of freedom. Here MS is 0.000166945.
 F (F-statistic): The F-statistic is a ratio of the variance explained by MS
Regression to the variance MS Residual. Here F is 0.092952483.
 Significance F (p-value): This is the p-value associated with the F-statistic,
Which is testing the null hypothesis means all the regression coefficients are
equivalent to zero. A high p-value (0.769321884) suggests that there is not
sufficient Proof to reject the null hypothesis.
2. Residual (Error):
 df (Degrees of Freedom): The df associated with the residuals or error term is
7, showing the number of observations minus the number of parameters
estimated in the model.
 SS (Sum of Squares): The sum of squared distinctness between the observed
values with the predicted values. Here SS is 0.012572173.
 MS (Mean Square): The mean square for the residuals, Done through SS
divided by the degrees of freedom. MS is 0.001796025.
3. Total:
 df (Degrees of Freedom): The total degrees of freedom, which is the combined
degrees of freedom for the regression and residuals. Here df is 8.
 SS (Sum of Squares): The total sum of squares, showing the total variability in
the dependent variable. SS = 0.012739118.
Interpretation:
 Regression (Model):

22 | P a g e
Md. Khaleduzzaman Monayam Hossain Moin Serajul Islam
20231164 20231160 20231167

 The F-statistic tests the general significance of the regression model. In this
case, since the F-statistic is not remarkably different from 1 , there is little
proof to reject the null hypothesis that the regression model has little
descriptive power.
 Residual (Error):
 The residual sum of squares (SS) calculates the inexplainable variability in the
dependent variable. A negligible residual sum of squares means a better fit of
the model to the data.
 Total:
 The total sum of squares means that the overall variability in the dependent
variable. It is the sum of the regression and residual sums of squares.
In summary, based on the ANOVA table, the simple linear regression model does not appear
to have a statistically remarkable overall effect, as indicated by the non-significant p-value
associated with the F-statistic. This represents that the model may not be giving a relevant
description of the variability in the dependent variable.
Coefficients Standard Error t Stat P-value Lower 95% Upper 95% Lower 95.0% Upper 95.0%
Intercept 0.0007239 0.025912529 0.027935291 0.9784935 -0.060549521 0.061997269 -0.060549521 0.061997269
Interest Rate (%) -0.0035736 0.011721194 -0.304881097 0.769321884 -0.03128979 0.024142649 -0.03128979 0.024142649

Intercept:
 Intercept Coefficient (0.000723874): The intercept is the output of the dependent
variable when the independent variable is zero. Here, when the independent variable
(X) is zero, the predicted value of the dependent variable (Y) is approximately
0.000723874. This value is the y-intercept of the regression line.
 Standard Error (0.025912529): The standard error is a estimate of the variability or
precision of the estimated intercept.
 t Stat (0.027935291): The t-statistic is the coefficient divided by its standard error. It
shows how many standard errors the coefficient is away from zero.
 P-value (0.9784935): The p-value related with the intercept check the null hypothesis
that the intercept is equal to zero. A high p-value means that the intercept may not be
notably different from zero.
 95% Confidence Interval (Lower 95% = -0.060549521, Upper 95% = 0.061997269):
This gives a scale within which we can be logically confident that the true population
parameter (intercept) lies.
Interest Rate (%):
 Interest Rate Coefficient (-0.00357357): In our case, for every one-unit rise in the
independent variable (X), the predicted value of the dependent variable (Y) goes
down by approximately 0.00357357. The negative sign indicates a negative relation
between the independent and dependent variables.
23 | P a g e
Md. Khaleduzzaman Monayam Hossain Moin Serajul Islam
20231164 20231160 20231167

 Standard Error (0.011721194): The standard error is a measure of the variability of the
estimated coefficient.
 t Stat (-0.304881097): The t-statistic is the coefficient divided by its standard error. It
indicates how many standard errors the coefficient is away from zero.
 P-value (0.769321884): The p-value associated with the coefficient for Interest Rate
tests the null hypothesis that this coefficient is equal to zero. A high p-value suggests
that this coefficient may not be significantly different from zero.
 95% Confidence Interval (Lower 95% = -0.03128979, Upper 95% = 0.024142649):
This provides a scale within which we can be logically confident that the true
population parameter for the Interest Rate coefficient lies.
Interpretation:

 The equation for the regression line is Y^=Intercept+(Slope×X), Substituting the

values: Y^=0.000723874−(0.00357357×X). This equation narrates the linear
relationship within the independent variable (X) and the predicted values of the
dependent variable (Y). The intercept is the predicted value of Y when X is zero. The
slope is the rate of change in Y for a unit change in X.
 For the Intercept, the p-value (0.9784935) is high, showing that the intercept may not
be remarkably different from zero. The 95% confidence interval includes zero,
strengthening this explanation.
 For the Interest Rate coefficient, the p-value (0.769321884) is high, suggesting that
the coefficient may not be remarkably different from zero. The 95% confidence
interval as well add zero.
In summary, based on the findings, there isn't significant evidence to show that either the
intercept or the coefficient for Interest Rate is significantly different from zero.

Telecommunication industry

24 | P a g e
Md. Khaleduzzaman Monayam Hossain Moin Serajul Islam
20231164 20231160 20231167

Regression Statistics
Multiple R 0.05975574
R Square 0.003570748
Adjusted R Square -0.138776288
Standard Error 0.219794032
Observations 9

1. Multiple R (0.05975574):
 Multiple R, which is also known as the correlation coefficient, calculates the
solidity and path of the linear relationship between the independent variable
and the dependent variable. In this case, the value is approximately 0.0598,
meaning a very weak positive correlation.
2. R Square (0.003570748):
 R Square, which presents the proportion of the variance in the dependent
variable that is described by the independent variables. Here, only roughly
0.36% of the variability in the dependent variable is explained by the
independent variable.
3. Adjusted R Square (-0.138776288):
 In this case, the Adjusted R Square is approximately -0.1388. A negative
Adjusted R Square is not common and represents that the model may not be
well-fitted to the data or that there may be issues such as overfitting or
including unnecessary variables.
4. Standard Error (0.219794032):
 Standard Error is used to measure the average distance between the observed
values and the values predicted by the model. In this case, it is high at
approximately 0.2198.
5. Observations (9):
 The number of observations used in the regression analysis is 9 .
Interpretation:
 The Multiple R value (0.0598) shows a very weak positive linear relationship between
the independent variables and the dependent variable. The firmness of this
relationship is extremely low.
 The R Square value (0.00357) indicates that only a very small amount of variability in
the dependent variable is described by the independent variables. This advice that the
model may not be giving a good fit to the data.
 The negative Adjusted R Square value (-0.1388) is uncommon

25 | P a g e
Md. Khaleduzzaman Monayam Hossain Moin Serajul Islam
20231164 20231160 20231167

 The Standard Error (0.2198) is relatively high, indicating that the model's predictions
may have a small amount of mistakes.
In summary, In accordance with statistics, the simple linear regression model does not seem
to be a strongly fit for the data, and the independent variables may not be describing much of
the variability in the dependent variable.
ANOVA
df SS MS F Significance F
Regression 1 0.001212 0.001212 0.025085 0.878628258
Residual 7 0.338166 0.048309
Total 8 0.339378

1. Regression:
 df (Degrees of Freedom): The degrees of freedom related with the regression
model. In this case, df is 1, indicating the number of independent variables in
the model.
 SS (Sum of Squares): The sum of squared distinct the predicted values and
the mean of the dependent variable. In our case, SS is 0.001211833.
 MS (Mean Square): The mean square is calculated as SS divided by the
degrees of freedom. MS = 0.001211833.
 F (F-statistic): The F-statistic is a ratio of the variance explained by MS
Regression to the variance not explained by the MS Residual. In our case, F =
0.025084810.
 Significance F (p-value): This is the p-value related with the F-statistic, testing
the null hypothesis that all the regression coefficients are equal to zero. A high
p-value (0.878628258) means that there is insufficient evidence to reject the
null hypothesis.
2. Residual (Error):
 df (Degrees of Freedom): The degrees of freedom related with the residuals or
error term. here, df is 7, showing the number of observations minus the
number of parameters estimated in the model.
 SS (Sum of Squares): The sum of squared differences between the observed
values and the predicted values. In our case, SS equals 0.338165916.
 MS (Mean Square): The mean square for the residuals, calculated as SS
divided by the degrees of freedom. In this case MS is 0.004830942.
3. Total:
 df (Degrees of Freedom): The total degrees of freedom, which is the sum of
the degrees of freedom for the regression and residuals. In this case, df is 8.

26 | P a g e
Md. Khaleduzzaman Monayam Hossain Moin Serajul Islam
20231164 20231160 20231167

 SS (Sum of Squares): The total sum of squares, which representing the general
variability in the dependent variable. In our case, SS equals to 0.339377749.
Interpretation:
 Regression (Model):
 In this case, As the F-statistic is not significantly different from 1, there is not
enough proof to reject the null hypothesis that the regression model has little
explanatory power.
 Residual (Error):
 A negligible residual sum of squares shows a better fit of the model to the
data.
 Total:
 The total sum of squares indicates the overall variability in the dependent
variable. It is the sum of the regression and residual sums of squares.
To sum up, based on the provided ANOVA table, the simple linear regression model does not
appear to have a statistically remarkable overall effect, as indicated by the non-significant p-
value associated with the F-statistic. This suggests that the model may not be providing a
strong relevant explanation of the variability in the dependent variable.

Coefficients Standard Error t Stat P-value Lower 95% Upper 95%Lower 95.0% Upper 95.0%
Intercept 0.105860021 0.134390801 0.787703 0.45671 -0.2119237 0.423644 -0.211924 0.423643769
Interest Rate (%) 0.00962802 0.060789922 0.158382 0.878628 -0.1341173 0.153373 -0.134117 0.153373345

Intercept:
 Intercept Coefficient (0.105860021): This is the estimated value of Y when X is zero.
The intercept is about 0.1058. This indicates that when X is zero, the estimated value
of Y is 0.1058.
 Standard Error (0.134390801): The standard error, which is a tool measure of the
variability or precision of the estimated intercept.
 t Stat (0.78770288): The t-statistic is the coefficient which is divided by its standard
error. It indicates how many standard errors the coefficient is away from zero.
 P-value (0.456709763): The p-value connected with the intercept tests the null
hypothesis that the intercept is equal to zero. A high p-value (more than 0.05) suggests
that there is not sufficient proof to reject the null hypothesis.
 95% Confidence Interval (Lower 95% = -0.211923727, Upper 95% = 0.423643769):
This provides a rang within which we can be logically confident that the true
population parameter (intercept) lies.
Interest Rate (%):

27 | P a g e
Md. Khaleduzzaman Monayam Hossain Moin Serajul Islam
20231164 20231160 20231167

 Interest Rate Coefficient (0.00962802): This represents the change in Y for a one-unit
change in X. Means that every one-unit rise in the independent variable X, the
dependent variable Y is expected to rise by approximately 0.0096 units. Conversely,
for every one unit down in X, Y is expected to decrease by the same amount.
 Standard Error (0.060789922): The standard error is a measure of the variability or
precision of the estimated coefficient.
 t Stat (0.15838185): The t-statistic is the coefficient divided by its standard error. It
indicates how many standard errors the coefficient is away from zero.
 P-value (0.878628258): The p-value associated with the coefficient for "Interest Rate
(%)" tests the null hypothesis that this coefficient is equal to zero. A high p-value
(greater than 0.05) recommends that there is not insufficient proof to reject the null
hypothesis.
 95% Confidence Interval (Lower 95% = -0.134117304, Upper 95% = 0.153373345):
This gives a range within which we can be reasonably confident that the true
population parameter for the "Interest Rate (%)" coefficient lies.
Interpretation:

 The regression equation Y=0.105860021+0.00962802⋅X , This equation make

predictions for Y based on different values of X
 For the Intercept, the p-value (0.456709763) is high, which suggests that the intercept
may not be remarkably different from zero. The 95% confidence interval includes
zero, reinforcing this interpretation.
 For the Interest Rate coefficient, the p-value (0.878628258) is high, suggesting that
the coefficient may not be highly different from zero. The 95% confidence interval
also includes zero.
 The t-stats for both intercept and the Interest Rate coefficient are low, further
indicating that these coefficients are not significantly different from zero.
To sum up, based on the provided results, there isn't strong proof to suggest that either the
intercept or the coefficient for Interest Rate is significantly different from zero.

Food Industry

28 | P a g e
Md. Khaleduzzaman Monayam Hossain Moin Serajul Islam
20231164 20231160 20231167

Regression Statistics
Multiple R 0.412962326
R Square 0.170537883
Adjusted R Square 0.052043295
Standard Error 0.544227304
Observations 9

1. Multiple R (0.412962326):
 Here, the value is nearly 0.413, which indicates a moderate positive
correlation.
2. R Square (0.170537883):
 In our case, about 17.05% of the variability in the dependent variable is
described by the independent variable.
3. Adjusted R Square (0.052043295):
 It is the adjusted R Square which is a modified version of R Square that
adjusts for the number of independent variables in the model. Here, the
adjusted R Square is lower than the R Square, which may indicate that some
of the independent variable's explanatory power is due to chance.
4. Standard Error (0.544227304):
 Standard Error is used to measure the average distance between the observed
values and the values predicted by the model. In this case, it is approximately
0.5442.
5. Observations (9):
 The number of data points used in the regression analysis. Here, there are 9
observations.
Interpretation:
 The Multiple R value (0.413) suggests a moderate positive linear relationship between
the dependent variable and the independent variable. In this case, strength of this
relationship is moderate.
 The R Square value (0.1705) shows that approximately 17.05% of the variability in
the dependent variable is explained by the independent variable. This implies a
moderate level of explanatory power.
 The Adjusted R Square value (0.0520) is lower compared to the R Square indicating
that some of the apparent explanatory power is due to chance or noise in the data.
 The Standard Error (0.5442) gives a measure of the average error in the expectations.
A lower standard error shows a more precise model.

29 | P a g e
Md. Khaleduzzaman Monayam Hossain Moin Serajul Islam
20231164 20231160 20231167

To sum up, according to the supplied statistics, the simple linear regression model appears to
have a moderate level of explanatory power, where the adjusted R Square suggests caution,
ANOVA
df SS MS F Significance F
Regression 1 0.426268268 0.426268 1.439204 0.269292818
Residual 7 2.073283507 0.296183
Total 8 2.499551775

The ANOVA (Analysis of Variance) table

[Link]:
 df (Degrees of Freedom): The degrees of freedom connected with the
regression model. Here, df = 1, indicating the number of independent variables
in the model.
 SS (Sum of Squares): The sum of squared distinctions between the anticipated
values and the mean of the dependent variable. In this case, SS =
0.426268268.
 MS (Mean Square): The mean square, which is calculated as SS divided by the
degrees of freedom. MS = 0.426268268.
 F (F-statistic): The F-statistic is a ratio of the variance explained by the MS
Regression to the variance not explained by MS Residual. In this case, F
equals to 1.43920398.
 Significance F (p-value): This is the p-value connected with the F-statistic,
which test the null hypothesis that all the regression coefficients are equal to
zero. A high p-value (0.269292818) means that there is insufficient evidence
to reject the null hypothesis.
2. Residual (Error):
 df (Degrees of Freedom): The degrees of freedom connected with the residuals
or error term. In this case, df = 7, indicating the number of observations minus
the number of parameters estimated in the model.
 SS (Sum of Squares): The sum of squared differences between the observed
values and the predicted values. In our case, SS equals to 2.073283507.
 MS (Mean Square): The mean square for the residuals is calculated as SS
divided by the degrees of freedom. Here, MS = 0.296183358.
3. Total:
 df (Degrees of Freedom): which is the sum of the degrees of freedom for the
regression and residuals. Here, df = 8.
 SS (Sum of Squares): The total sum of squares, which represents the general
variability in the dependent variable. In our case, SS = 2.499551775.

30 | P a g e
Md. Khaleduzzaman Monayam Hossain Moin Serajul Islam
20231164 20231160 20231167

Interpretation:
 Regression (Model):
 F-statistic tests the overall importance of the regression model. In this case, as
the F-statistic is not notably different from 1, there is limited evidence to reject
the null hypothesis that the regression model has little explanatory power.
 Residual (Error):
 The residual sum of squares (SS) that measures the unexplained variability in
the dependent variable. A negligible residual sum of squares reflects a better
fit of the model to the data.
 Total:
 The total sum of squares indicates the overall variability in the dependent
variable. This is the sum of the regression and residual sums of squares.
To sum up, based on the provided ANOVA table, the simple linear regression model does not
appear to have a statistically remarkable overall effect, which is indicated by the non-
significant p-value associated with the F-statistic. It suggests that the model may not be
providing a strong meaningful justification of the variability in the dependent variable.
Coefficients Standard Error t Stat P-value Lower 95% Upper 95% Lower 95.0% Upper 95.0%
Intercept -0.39868394 0.332762189 -1.1981 0.269863 -1.185541485 0.3881736 -1.1855415 0.3881736
Interest Rate (%) -0.18057482 0.150520627 -1.19967 0.269293 -0.536499546 0.1753499 -0.5364995 0.1753499

Intercept:
 Intercept Coefficient (-0.398683942): The intercept is the value of Y while X is zero.
The intercept is about -0.399. This indicates that when X is zero, the estimated value
of Y is -0.399.
 Standard Error (0.332762189): The standard error, which is a tool to measure the
variability of the approximated intercept.
 t Stat (-1.198104699): The t-statistic is the coefficient which is divided by its standard
error. It indicates how many standard errors the coefficient is away from zero.
 P-value (0.26986321): The p-value connected with the intercept tests the null
hypothesis that the intercept is equal to zero. A high p-value (more than 0.05) suggests
that there is not enough proof to reject the null hypothesis.
 95% Confidence Interval (Lower 95% = -1.185541485, Upper 95% = 0.3881736): It
provides a range within which it can be reasonably confident that the true data
parameter lies.
Interest Rate (%):
 Interest Rate Coefficient (-0.180574821): This represents the change in Y for a one-
unit change in X. Here, it is approximately -0.181, which implies that for every one-

31 | P a g e
Md. Khaleduzzaman Monayam Hossain Moin Serajul Islam
20231164 20231160 20231167

unit increase in X, the estimated value of Y decreases by 0.181 units. In contrast, for
every one-unit decrease in X, the estimated value of Y increases by 0.181 units.
 Standard Error (0.150520627): The standard error is a tool that measures the
variability or precision of the approximated coefficient.
 t Stat (-1.199668279): The t-statistic is the coefficient divided by its standard error,
which indicates how many standard errors the coefficient is away from zero.
 P-value (0.269292818): The p-value associated with the coefficient for Interest Rate
tests the null hypothesis that the coefficient is equal to zero. A high p-value (more
than 0.05) recommends that there is insufficient evidence to reject the null hypothesis.
 95% Confidence Interval (Lower 95% = -0.536499546, Upper 95% = 0.175349903):
This gives a range in which we can be reasonably confident that the true population
parameter for the Interest Rate coefficient lies.
Interpretation:

 the regression equation Y=−0.398683942−0.180574821⋅X . The equation makes

predictions for Y based on different values of X.
 For the Intercept, the p-value (0.26986321) is high, which suggests that the intercept
may not be significantly different from zero. The 95% confidence interval includes
zero.
 For the Interest Rate coefficient, the p-value (0.269292818) is high, which suggests
that the coefficient may not be significantly distinct from zero. Here, 95% confidence
interval also includes zero.
 The t-stats for both the Intercept and the Interest Rate coefficient are small, which
further indicates that these coefficients are not significantly different from zero.
 To sum up, there isn't strong evidence to suggest that either the intercept or the
coefficient for Interest Rate is significantly different from zero.

32 | P a g e
Md. Khaleduzzaman Monayam Hossain Moin Serajul Islam
20231164 20231160 20231167

Diagnostics
Jarque Bera Test
The Jarque-Bera test is a test which is used to ascertain if a dataset has skewness and kurtosis
similar to a normal distribution
Oil Industry

Skewness 1.173057717
Excel Kurtosis 2.474093654
Kurtosis 5.474093654

JB 4.35952389
Jb 4.35952389
CV 1.386294361
P 0.113068444

Skewness: It measures the imbalance of the distribution. Here, the skewness is approximately
1.17, which indicates a moderate positive skew.
Kurtosis: Kurtosis is usually used to measure the tail heaviness of a distribution in contrast to
a normal distribution. Kurtosis of 3 is considered normal; values more than 3 indicate heavier
tails. The kurtosis is about 5.47, which indicates heavier tails than a normal distribution.
JB / Jb: The value of the Jarque-Bera test statistic in our case, it's nearly 4.36.
CV: Critical value at a certain significance level (5% or 1%). For this dataset, the critical
value is nearly 1.39.
P: P-value related to the Jarque-Bera test statistic. It is approximately 0.113.
Interpretation:
If P is smaller than a chosen significance level (commonly 0.05), we will reject the null
hypothesis. It means that here is enough information to say that the data is not normally
distributed.
If P is greater than the significance stage, we will accept the null hypothesis, which refers that
there is not enough information to conclude that the data significantly deviates from a normal
distribution.
In our case, with a p-value of approximately 0.113, which is greater than 0.05, we would fail
to reject the null hypothesis at the significance level of 0.05. So, based on this test, we don't
have enough evidence to say that the data deviates from a normal distribution. Considering

33 | P a g e
Md. Khaleduzzaman Monayam Hossain Moin Serajul Islam
20231164 20231160 20231167

the value of p is relatively close to 0.05, there might be some indication of deviation, but it's
not strong enough to make a definitive conclusion.
Telecommunication Industry

Skewness 2.02219
Excel 6.67826
Kurtosis 7
9.67826
Kurtosis 7

JB 22.8586
Jb 22.8586
1.38629
CV 4
1.09E-
P 05

Skewness: Here a positive skewness (like ours, at 2.022) implies that given data is skewed to
right.
Excel Kurtosis: It measures the "peakedness" of distribution compared to a normal
distribution. A value more than 3 indicates heavier tails than a normal distribution.
Kurtosis: It is another measure of the shape distribution. The higher kurtosis value (9.678)
indicates a distribution with heavier tails and a sharper peak in contrast to a normal
distribution.
JB (Jarque-Bera): This is the test result. In our case, it's 22.8585998.
CV (Critical Value): This is critical value at a certain significance level. Data scientists use it
to ascertain whether the test statistic is within an acceptable range.
P (p-value): This is the probability of noticing the test if null hypothesis is true. Here a very
tiny p-value (like 1.08722E-05, which is essentially 0) shows strong evidence against null
hypothesis, meaning this data significantly comes from the normal distribution.
Here, since the p-value is very low, we will reject the null hypothesis. This data doesn't seem
to follow a normal distribution based on its skewness and kurtosis, which is indicated by
Jarque-Bera test.
Food Industry
Skewness -0.49103
Excel Kurtosis -0.74819
2.25180
Kurtosis 7

34 | P a g e
Md. Khaleduzzaman Monayam Hossain Moin Serajul Islam
20231164 20231160 20231167

0.57159
JB 1
0.57159
Jb 1
1.38629
CV 4
0.75141
P 6

Skewness: A negative skewness of -0.491 refers a small leftward skewness (tail to the left) in
the data.
Excel Kurtosis: This is an Excel calculation or representation of kurtosis (-0.748).
Kurtosis: It measures whether a distribution is peaked or flat. If the kurtosis is 2.25, it
indicates that the distribution is more peaked, compared to when it is normal.
As example, p-value is 0.751 and the JB statistic of 0.571 may not be enough to reject null
hypothesis that the provided information is distributed normally because the value of p is
higher than the typical significance values of 0.05. So, taking into account the data's
skewness, kurtosis, and the Jarque-Bera test results, we could erroneously claim that it might
follow a normal distribution based just on this test.

Heteroskedasticity White Test

The White test is a test that is used to detect heteroskedasticity in a regression analysis.
Heteroskedasticity is the situation where the variability errors (or residuals) in a regression
model are not constant. Simply, this means that the residual spread is not consistent with the
predicted values.

Oil Industry
R Square 0.013104907
LM 0.117944164

R Square: It is the R-squared value from a regression model. This is used to measure the
proportion of the variance in a dependent variable which is dependable from the independent
variables. Here, it indicates how well the independent variables explain the variability in the
dependent variable. The value of 0.0131 suggests that independent variables clarify
approximately 1.31% of the variance in dependent variable.
LM: This is the test resulting from the White test for heteroskedasticity. In our case, the test
statistic is 0. [Link] importance of the LM helps ascertain whether there's evidence to not

35 | P a g e
Md. Khaleduzzaman Monayam Hossain Moin Serajul Islam
20231164 20231160 20231167

accept the null hypothesis of homoskedasticity in favor of the heteroskedasticity alternative

theory.
Without the critical value or further context, it is impossible to definitively determine whether
the test result says that there is heteroskedasticity. The test statistic itself (0.1179) doesn't
provide a direct indication of significance without comparison to a critical value.
Telecommunication Industry
R Square 0.003571
LM 0.032137
R Square: 0.003570748 is the value in our case. Here, the model explains a very small
portion of the variance in the dependent variable.
LM: 0.032136736 (This might be the test statistic value gained from conducting the White
test.)
Without additional context or the critical value for comparison, it's challenging to interpret
the result. In the White test, if the test statistic exceeds the chi-squared distribution's critical
value with a certain degree of freedom, it shows there is heteroskedasticity.
Here, with the given LM of 0.032136736, it seems small. Normally, a higher test statistic
would suggest the presence of heteroskedasticity, but without the critical value or degrees of
freedom, it's not possible to ascertain the significance of the test result.
The interpretation would largely depend on compare test statistic against the value and the
associated significance level to find out the presence of heteroskedasticity.
Food Industry
R Square 0.170537883
LM 1.534840946

R Square: This value is 0.1705, and represents the coefficient of determination of our model.
R-squared value is 0.1705, which means that approximately 17.05% of the variance in the
dependent variable.
LM: The LM resulting from conducting the White test for heteroskedasticity. In our case, the
LM is 1.5348.
The test statistics follow a chi-square distribution. To ascertain is there any evidence of
heteroskedasticity, this LM needs to be compared against a critical value from the chi-square
distribution. The critical value relies on the significant level selected for the test and the
degrees of [Link] the LM is greater compared to the critical value, it refers to the
evidence of heteroskedasticity in the regression model. If it's smaller, it suggests that there
might not be significant heteroskedasticity.

Goodness of fit Test

36 | P a g e
Md. Khaleduzzaman Monayam Hossain Moin Serajul Islam
20231164 20231160 20231167

Oil Industry
R Square 0.013104907
In our case, R-squared value is [Link] is used to measure the goodness of fit of a
regression model.
R-squared is also called coefficient of determination, which determines the percentage of the
dependent variable's volatility that is explained by the independent variable in a regression
model.
In our case, the value of R-squared is 0.0131, which suggests that only 1.31% variability in
the dependent variable is explained by the independent variable.
A low R-squared value implies that the model does not properly fit with provided data. This
implies that the independent variable does not clarify the variability observed in the
dependent variable.
It's crucial to take into account the context of our analysis when explaining the R-squared
value. Real-world data may have inherent variability which may not be fully outlined by the
variables used in our model.
Telecommunication Industry

R Square 0.003571

In the telecommunication industry, R-squared value is 0.003570748.

Here, an R-squared value of approximately 0.0035 suggests that the dependent variable is
explained by the independent variable only about 0.35%.
A lower R-squared value means the model doesn't effectively elucidate the variations of the
dependent variable. This may imply that the chosen independent variables might not be
strongly related to or predictive of the dependent variable. This low value could also suggest
that there might be other variables or factors that are influencing the dependent variable.
An R-squared of 0.0035 refers that the model might not be a good representation of the
connection between the variables under study.

Food Industry
R Square 0.170538

The result of a Goodness of Fit test, indicated by the value of R-squared which is
0.170537883, shows the percentage of the dependent variable's volatility that is accounted for
by the independent variable.
The value of R-squared between 0 and 1, there

37 | P a g e
Md. Khaleduzzaman Monayam Hossain Moin Serajul Islam
20231164 20231160 20231167

- 0 means it does not clarify any of the response data's fluctuation around its mean.
- 1 shows that all the response data variability around its mean.
In our case, R-squared value is nearly 0.17 (17%), which implies that the independent
variable in this model can explain about 17% of the variability observed in the dependent
variable. It suggests that while there is a connection between the variables, a significant
portion of the dependent variable remains unexplained in this model.
It is essential to consider the specific field of study when interpreting the goodness of fit. In
some studies, an R-squared value is 0.17, which may be considered good, while in other
scenarios, that may be considered insufficient for accurate predictions.

Conclusion
The overall purpose of the paper has been to analyze the impact of interest rates on the three
(03) different industries. And the sample contains 15 companies. The model consists of the
intercept and slope. In our model, we have observed that the telecommunication industry has
the highest intercept. Since we have a negative coefficient in the food industry, it tells us that
when the interest rate is zero the food industry has negative growth. And other two industries
have a positive coefficient.
We observed that in the food industry even though the interest rate is zero the investment
growth rate is minus, which means they are having a very bad time maybe because of the
coronavirus pandemic effect. Other than that, the oil & food industry has a positive result
even though the interest rate is zero. Another thing is that the food industry investment may
be highly leveraged. That’s why maybe the coefficient is negative. If we took more
companies and more periods or excluded the pandemic period, we could find a more accurate
desired result. We have conducted several tests which result show similar findings. In our
model in Jarque Bera test the food industry is lower value where the telecommunication
industry is highest value, which refers to the high deviation from the normal distribution. In
the white thet the value of 03 industry refers that all have heteroskedasticity.
In navigating the complex interplay of U.S. interest rates and industry-specific investments,
our study provides valuable insights for investors, policymakers, and industry professionals.
As financial landscapes evolve, understanding these dynamics becomes paramount for
making informed decisions and staying resilient in a dynamic economic environment. This
study contributes to the ongoing discourse on the intricate connections between
macroeconomic variables and corporate investment strategies.

38 | P a g e
Md. Khaleduzzaman Monayam Hossain Moin Serajul Islam
20231164 20231160 20231167

Literature list
Medium
[Link]
b3f5528e54b7#:~:text=Here%20is%20the%20formula%3A,be%20predicted%20based
%20on%20Age.
Statistics how to
[Link]
XLSTAT
[Link]
Verizon :
[Link]
T-Mobile :
[Link]
Comcast Crop:
[Link]
AT&T :
[Link]
charter communications :
[Link]
Campbell Soup
[Link]
General Mills
[Link]

39 | P a g e
Md. Khaleduzzaman Monayam Hossain Moin Serajul Islam
20231164 20231160 20231167

Mondelez International
[Link]
Pepsico
[Link]
Conagra
[Link]
Chevron
[Link]
Exxon Mobil
[Link]
ConocoPhillips
[Link]
Marathon Petroleum
[Link]
Valero Energy
[Link]

40 | P a g e