Business Analytics and AI Application Term Project

Regression Analysis of Financial Drivers for a Listed Company

Project Title

Regression Analysis: What Drives Stock Price Changes for [Selected Company]?

Objective

Apply regression analysis to real annual financial data to discover which financial metrics most impact the stock price of a publicly listed company. The project bridges statistical analysis with real-world business decision-making.

Project Workflow

1. Company & Question Selection

Each student selects a publicly listed company in their home country.

Step-by-Step Instructions: Selecting a Publicly Listed Company

Here are concrete steps for students as the first phase of your MBA regression analysis project:

1. Brainstorm and Shortlist

List 3–5 companies from your home country’s stock exchange that interest you.
Consider industries you want to learn about (technology, finance, manufacturing, etc.).

2. Confirm Public Listing Status

Search for each company on your country’s main stock exchange website (e.g., KOSPI, NSE, TSE, etc.).
Make sure the company is actively traded and regularly publishes annual reports.

3. Check Data Availability

Visit the investor relations section of each company’s official website.
Confirm that you can access at least 5–10 years of annual financial statements (income statement, balance sheet, cash flow).
Check if historical stock price data is available for free.

4. Create a Short Rationale

For your top company, write 2–3 sentences explaining why you chose it (personal interest, industry relevance, notable recent changes, etc.).

Example (for Samsung Electronics, Korea)

Company: Samsung Electronics Co., Ltd.

Stock Symbol: 005930 (KOSPI)

Reason for Selection: Samsung is Korea’s leading technology company with a robust historical dataset and global relevance. I am interested in exploring what drives its stock price given recent shifts in the semiconductor industry.

Investor Relations Link: https://www.samsung.com/global/ir/

Research Question

Define a clear business question, e.g.: “What factors most strongly predict changes in [Company]’s stock price?”

2. Data Acquisition

Purpose

Gather comprehensive historical financial and stock market data that forms the foundation of your regression analysis. Quality data acquisition ensures reliable and valid analytical results.

Metrics

Collect the following data for at least 5–10 consecutive years:

Financial Statement Data:

Income Statement: Revenue, Net Income, Operating Income, EBITDA, R&D Expenses, Marketing Expenses
Balance Sheet: Total Assets, Total Liabilities, Shareholders’ Equity, Current Assets, Current Liabilities
Cash Flow Statement: Operating Cash Flow, Investing Cash Flow, Financing Cash Flow, Free Cash Flow

Market Data:

Stock Prices: Annual closing prices, adjusted for stock splits and dividends
Market Benchmark: Relevant market index values (e.g., KOSPI, S&P 500, Nikkei 225)
Trading Volume: Annual average trading volume (optional)

Calculated Ratios (if not directly available):

Earnings Per Share (EPS)
Debt-to-Equity Ratio
Return on Equity (ROE)
Return on Assets (ROA)
Price-to-Earnings Ratio (P/E)

Expected Output

A comprehensive dataset with the following characteristics:

Time Coverage: Minimum 5 years, preferably 10+ years of annual data
Data Sources: Official company filings (10-K, annual reports), financial databases
Format: Organized spreadsheet (Excel or CSV) with:
- Rows: Years (one row per year)
- Columns: All financial metrics and stock prices
- Clear labeling with units (e.g., millions, billions)
Documentation: Note the source and date of data collection for each metric
Completeness: All selected variables should have data for all years (minimal missing values)

3. Dataset Construction

Purpose

Transform raw financial data into a structured analytical dataset with clearly defined dependent and independent variables. This step establishes the foundation for your regression model by identifying what you want to predict and which factors might drive that outcome.

Metrics

Dependent Variable (Target/Outcome):

Choose ONE of the following as your primary dependent variable:

Stock Price: Annual end-of-year closing price (adjusted for splits)
Stock Price Change (%): Year-over-year percentage change in stock price
Stock Return: (Price_t - Price_t-1) / Price_t-1
Market-Adjusted Return: Stock return minus market index return

Independent Variables (Predictors):

Select 5-10 relevant financial metrics from the following categories:

Profitability Metrics:

Revenue (or Revenue Growth %)
Net Income
Earnings Per Share (EPS)
Operating Margin (%)
Return on Equity (ROE)
Return on Assets (ROA)

Financial Health Metrics:

Debt-to-Equity Ratio
Current Ratio
Total Assets
Total Debt

Investment & Growth Metrics:

R&D Expense (or R&D as % of Revenue)
Capital Expenditure
Free Cash Flow
Operating Cash Flow

Market Context Variables:

Market Index Return (control variable)
Industry-specific benchmark (optional)

Expected Output

A clean, analysis-ready dataset structured as follows:

Table Structure:

Year	Stock_Price	Revenue	Net_Income	EPS	Debt_to_Equity	RD_Expense	ROE	Market_Return
2015	185.50	95.2B	7.2B	3.10	1.45	1.8B	12.5	5.2
2016	198.20	102.1B	7.8B	3.25	1.52	1.9B	13.1	6.8
2017	210.75	108.5B	8.5B	3.45	1.48	2.1B	14.2	8.1
…	…	…	…	…	…	…	…	…

Quality Checklist:

Consistent units across all years (e.g., all revenue in billions)
Variables aligned with business logic and theoretical relationships
All variables historically available (no look-ahead bias)
Clear variable naming conventions
Data dictionary documenting each variable’s definition and unit
Variables selected based on industry characteristics and research question

4. Data Verification & Preprocessing

Purpose

Ensure data quality and prepare the dataset for valid statistical analysis. This critical step identifies and corrects data issues that could compromise your regression results, including missing values, outliers, and inconsistencies.

Metrics

Data Quality Checks:

Completeness Assessment
- Missing value count per variable
- Percentage of missing data (< 10% acceptable, < 5% ideal)
- Pattern of missingness (random vs systematic)
Validity Checks
- Range checks (e.g., ratios within plausible bounds)
- Sign checks (e.g., revenue should be positive)
- Consistency checks (e.g., Net Income = Revenue - Expenses)
- Stock split adjustments verified
Outlier Detection
- Z-scores for continuous variables (|z| > 3 indicates potential outlier)
- Box plots for visual inspection
- Cook’s Distance for influential observations
- Historical context (e.g., merger, restructuring may explain outliers)
Distributional Analysis
- Skewness and kurtosis of variables
- Normality tests (Shapiro-Wilk, Q-Q plots)
- Identify variables requiring transformation

Preprocessing Steps:

Missing Data Handling:
- Document missing values and their causes
- Apply imputation if appropriate (mean, median, or forward-fill)
- Consider excluding variables with >15% missing data
- Justify your approach
Outlier Treatment:
- Verify outliers are not data entry errors
- Investigate business context (legitimate vs error)
- Options: keep (if valid), winsorize (cap at percentile), or remove
- Document decisions
Variable Transformation:
- Log transformation for right-skewed variables (revenue, assets)
- Percentage changes for growth rates
- Standardization (z-scores) if comparing variables on different scales
- Lag variables if testing time-delayed effects
Multicollinearity Check:
- Correlation matrix of independent variables
- Variance Inflation Factor (VIF) calculation
- VIF > 10 indicates severe multicollinearity (consider removing variables)

Expected Output

A clean, preprocessed dataset ready for regression analysis with:

Summary Statistics Table:

Variable	N	Mean	Std Dev	Min	Max	Missing	Outliers
Stock_Price	10	205.3	18.5	175.2	235.8	0	0
Revenue	10	102.5B	8.2B	90.1B	115.2B	0	0
EPS	10	3.42	0.28	3.05	3.85	1	0
Debt_to_Equity	10	1.52	0.15	1.28	1.75	0	1

Preprocessing Report:

Data quality assessment completed with documented findings
Missing values handled with justification (e.g., “Revenue 2018 imputed using linear interpolation”)
Outliers identified and addressed (e.g., “2020 Net Income outlier retained due to verified one-time gain”)
Variables transformed as needed (e.g., “Log(Revenue) created due to right skew”)
Stock splits and dividends adjusted in price data
Correlation matrix shows no severe multicollinearity (all r < 0.8)
Clean dataset saved with preprocessing log documenting all changes

5. Regression Analysis

Purpose

Build a statistical model to quantify the relationship between financial metrics and stock price, identifying which factors are significant predictors and estimating the magnitude of their effects. This transforms your research question into a testable mathematical model.

Metrics

Model Specification:

Simple Linear Regression (for preliminary analysis):

\[\text{Stock Price}_t = \beta_0 + \beta_1 \times \text{Predictor}_t + \epsilon_t\]

Example: Stock Price = β₀ + β₁ × EPS + ε

Multiple Linear Regression (primary model):

\[\text{Stock Price}_t = \beta_0 + \beta_1 X_1 + \beta_2 X_2 + ... + \beta_k X_k + \epsilon_t\]

Example:

Stock Price = β₀ + β₁(Revenue) + β₂(Net Income) + β₃(Debt/Equity)
              + β₄(R&D Expense) + β₅(Market Return) + ε

Key Regression Metrics to Report:

Coefficients (β)
- Interpretation: Change in dependent variable for one-unit change in independent variable
- Sign: Positive (+) or negative (-) relationship
- Magnitude: Strength of effect
- Example: β₁ = 15.5 means “Each $1B increase in revenue associates with $15.50 stock price increase”
Statistical Significance
- p-value: Probability the relationship is due to chance
  - p < 0.05: Statistically significant (conventional threshold)
  - p < 0.01: Highly significant
  - p ≥ 0.05: Not significant
- t-statistic: Coefficient divided by standard error (|t| > 2 generally significant)
Standard Error (SE)
- Measures uncertainty in coefficient estimate
- Smaller SE = more precise estimate
Confidence Intervals (95% CI)
- Range within which true coefficient likely falls
- Example: β₁ = 15.5, 95% CI [10.2, 20.8]

Model Selection Approach:

Start with bivariate regressions (one predictor at a time) to understand individual relationships
Build multiple regression with theory-driven variable selection
Consider stepwise regression or best subsets for exploratory purposes
Justify final model based on theory, significance, and model fit

Expected Output

Regression Results Table:

Variable	Coefficient (β)	Std Error	t-value	p-value	95% CI	Significance
(Intercept)	45.30	12.50	3.62	0.008	[18.2, 72.4]	**
Revenue (B)	1.25	0.35	3.57	0.009	[0.48, 2.02]	**
EPS	12.80	4.20	3.05	0.018	[3.45, 22.15]	*
Debt_to_Equity	-8.50	6.30	-1.35	0.219	[-22.1, 5.1]	NS
R&D_Expense(B)	3.40	1.80	1.89	0.098	[-0.75, 7.55]	†
Market_Return	2.15	0.85	2.53	0.036	[0.25, 4.05]	*

*Significance codes: ** p<0.01, * p<0.05, † p<0.10, NS not significant*

Model Summary:

R² (R-squared): Proportion of variance explained (e.g., R² = 0.75 means model explains 75% of stock price variation)
Adjusted R²: R² adjusted for number of predictors (penalizes overfitting)
F-statistic: Tests overall model significance (p < 0.05 indicates model is significant)
Root Mean Square Error (RMSE): Average prediction error in original units
Sample size: Number of observations (years) used

Interpretation Example:

“The multiple regression model explains 75% of the variation in stock price (Adj. R² = 0.72). Revenue and EPS are statistically significant positive predictors (p < 0.05). Specifically:

Each $1 billion increase in revenue is associated with a $1.25 increase in stock price (p = 0.009)
Each $1 increase in EPS is associated with a $12.80 increase in stock price (p = 0.018)
Market return has a moderate positive effect (β = 2.15, p = 0.036)
Debt-to-Equity ratio shows no significant relationship (p = 0.219)

These findings suggest that profitability metrics (Revenue, EPS) are the primary drivers of [Company]’s stock price.”

6. Model Evaluation

Purpose

Validate that your regression model is statistically sound and meets the required assumptions for valid inference. This step ensures your conclusions are reliable and that the model can be trusted for business decision-making.

Metrics

Model Fit Assessment:

R-squared (R²)
- Interpretation: Percentage of variance in dependent variable explained by the model
- Good: R² > 0.60 for time series financial data
- Acceptable: 0.40 ≤ R² < 0.60
- Weak: R² < 0.40 (model has limited explanatory power)
Adjusted R-squared
- Adjusts for number of predictors (penalizes overfitting)
- Should be close to R² (gap < 0.05)
- Use this for comparing models with different numbers of predictors
F-statistic
- Tests overall model significance (all coefficients = 0?)
- Requirement: p-value < 0.05 (model is statistically significant)
- Higher F-value indicates stronger overall model
Root Mean Square Error (RMSE)
- Average prediction error in original units (e.g., dollars)
- Lower is better
- Compare to standard deviation of dependent variable
- RMSE < 0.5 × SD indicates good fit

Regression Assumption Tests:

Linearity
- Test: Scatter plots of each predictor vs dependent variable
- Check: Residuals vs fitted values plot (should show no pattern)
- Requirement: Relationship should be approximately linear
- Violation: Non-random pattern suggests non-linear relationship
Normality of Residuals
- Test: Q-Q plot (quantile-quantile plot)
- Test: Histogram of residuals
- Test: Shapiro-Wilk test (p > 0.05 indicates normality)
- Requirement: Residuals should be approximately normally distributed
- Note: Less critical with larger sample sizes (n > 30)
Homoscedasticity (Constant Variance)
- Test: Residuals vs fitted values plot
- Test: Breusch-Pagan test (p > 0.05 indicates homoscedasticity)
- Requirement: Variance of residuals should be constant across all fitted values
- Violation: Funnel shape in residual plot
Independence of Errors
- Test: Durbin-Watson statistic
- Range: 1.5 to 2.5 is acceptable (2.0 is ideal)
- Issue: Autocorrelation common in time series data
- Note: Document if present; may need time series methods
Multicollinearity
- Test: Variance Inflation Factor (VIF) for each predictor
- Threshold: VIF < 5 (acceptable), VIF < 10 (maximum acceptable)
- Test: Correlation matrix (pairwise correlations < 0.80)
- Violation: High VIF inflates standard errors, making coefficients unreliable
Influential Observations
- Test: Cook’s Distance (D)
- Threshold: D > 1 indicates highly influential observation
- Test: Leverage values (high leverage points)
- Action: Investigate influential points; determine if they should be retained

Expected Output

Model Fit Summary:

Multiple R-squared: 0.752
Adjusted R-squared: 0.720
F-statistic: 23.5 on 5 and 4 DF, p-value: 0.003
RMSE: 8.42
Residual standard error: 9.15 on 4 degrees of freedom

Assumption Test Results Table:

Assumption	Test	Result	Interpretation	Status
Linearity	Residual vs Fitted	No pattern	Linear relationships	✓ Met
Normality	Shapiro-Wilk	p = 0.18	Residuals normal	✓ Met
Homoscedasticity	Breusch-Pagan	p = 0.42	Constant variance	✓ Met
Independence	Durbin-Watson	DW = 1.85	No autocorrelation	✓ Met
Multicollinearity	VIF	All < 5	Low multicollinearity	✓ Met
Influential Cases	Cook’s D	Max = 0.45	No influential outliers	✓ Met

VIF Table (Multicollinearity Check):

Variable	VIF	Status
Revenue	3.42	✓ OK
EPS	2.18	✓ OK
Debt_to_Equity	1.85	✓ OK
R&D_Expense	4.15	✓ OK
Market_Return	1.22	✓ OK

Diagnostic Plots:

Include the following visualizations:

Residuals vs Fitted Values: Check linearity and homoscedasticity
Q-Q Plot: Check normality of residuals
Scale-Location Plot: Check homoscedasticity
Residuals vs Leverage: Identify influential observations
Histogram of Residuals: Visual check of normality

Evaluation Checklist:

If Assumptions Are Violated:

Linearity violation: Try polynomial terms or log transformation
Normality violation: With n > 30, proceed with caution; consider robust regression
Heteroscedasticity: Use robust standard errors or weighted least squares
Multicollinearity: Remove correlated predictors or use dimension reduction
Influential outliers: Investigate and justify retention or removal
Autocorrelation: Consider time series models (ARIMA) or add lagged variables

7. Insights & Recommendations

Purpose

Translate statistical findings into actionable business insights and strategic recommendations. This step bridges the gap between analytical results and practical decision-making, demonstrating the real-world value of your regression analysis.

Metrics

Key Findings Synthesis:

Significant Predictors Identification
- List all statistically significant variables (p < 0.05)
- Rank by effect size (standardized coefficients or magnitude)
- Categorize: Strong drivers (p < 0.01), Moderate drivers (0.01 ≤ p < 0.05)
Effect Magnitude Interpretation
- Quantify practical significance (not just statistical significance)
- Translate coefficients into business-relevant terms
- Example: “A 10% increase in R&D spending is associated with a $2.35 increase in stock price”
Comparative Analysis
- Which financial metrics have the strongest impact?
- Which metrics are surprisingly non-significant?
- How do findings compare to industry norms or prior research?
Business Context Integration
- Connect findings to company strategy and market conditions
- Explain why certain relationships exist (or don’t exist)
- Consider industry-specific factors (e.g., R&D critical for tech, not for retail)

Stakeholder-Specific Insights:

For Investors: Which financial signals predict stock performance?
For Management: Which operational levers drive shareholder value?
For Analysts: How does this company compare to industry benchmarks?

Expected Output

Executive Summary of Findings:

“This regression analysis examined 10 years of financial data to identify key drivers of [Company]’s stock price. The model explains 75% of stock price variation (R² = 0.75, p < 0.001). Three financial metrics emerged as significant predictors:”

Key Findings Table:

Finding	Metric	Coefficient	p-value	Business Interpretation
1	EPS	β = 12.80	0.018	Strongest driver: Each $1 increase in EPS → $12.80 stock price increase
2	Revenue	β = 1.25	0.009	Revenue growth signals market expansion; $1B revenue → $1.25 price increase
3	Market Return	β = 2.15	0.036	Stock moves with overall market; 1% market gain → $2.15 price increase
4	R&D Expense	β = 3.40	0.098	Marginally significant; suggests market values innovation investment
5	Debt-to-Equity	β = -8.50	0.219	Not significant: Leverage doesn’t significantly impact stock price

Strategic Insights:

Profitability is Paramount
- EPS has the strongest effect on stock price
- Implication: Investors prioritize earnings quality over top-line growth
- Context: Aligns with mature company valuations focused on profitability
Revenue Growth Matters, But Less Than Earnings
- Revenue coefficient positive but smaller than EPS
- Implication: Growth is valued, but not at expense of profitability
- Context: Market rewards sustainable, profitable growth
Market Sentiment Drives Performance
- Positive market return coefficient confirms systematic risk
- Implication: Company performance tied to broader economic conditions
- Context: Beta analysis suggests moderate market sensitivity
Leverage Surprisingly Irrelevant
- Debt-to-Equity ratio not significant predictor
- Implication: Market may view current debt levels as optimal or manageable
- Context: Industry norm or company’s debt is within acceptable range

Actionable Recommendations:

For Company Management:

Prioritize Earnings Quality (EPS)
- Action: Focus on margin improvement and operational efficiency
- Expected Impact: 10% EPS improvement could boost stock price by ~$1.28
- Implementation: Cost optimization, pricing strategy, productivity initiatives
Maintain Revenue Growth Trajectory
- Action: Invest in market expansion and product development
- Expected Impact: $10B revenue increase → ~$12.50 stock price gain
- Implementation: Market penetration, new product launches, M&A
Consider Increasing R&D Investment
- Action: Boost R&D spending by 15-20% (currently marginally significant)
- Rationale: May strengthen innovation perception and future growth potential
- Risk: Monitor profitability impact; avoid diluting earnings
Optimize Capital Structure (Lower Priority)
- Action: Debt-to-Equity ratio is not a current concern
- Strategy: Maintain financial flexibility; use debt for value-creating investments
- Note: Monitor if market conditions change

For Investors:

Key Metrics to Monitor
- Primary: Quarterly EPS trends and guidance
- Secondary: Revenue growth rates and R&D pipeline
- Context: Market conditions and sector performance
Valuation Signals
- Strong EPS growth likely to drive price appreciation
- Revenue stagnation may not severely impact price if profitability maintained
- Market downturns will affect stock despite strong fundamentals
Risk Factors
- Model explains 75% of variation; 25% from other factors
- Market beta suggests moderate systematic risk
- Company-specific risks: competitive dynamics, regulation, technology shifts

Limitations and Caveats:

Sample Size
- Analysis based on 10 years of data (small sample for time series)
- Limited statistical power to detect weak effects
- Recommendation: Update analysis as more data becomes available
Model Limitations
- Linear model may not capture non-linear relationships
- Annual data masks quarterly volatility
- Historical relationships may not persist in changing market conditions
Omitted Variables
- Model doesn’t include: competitive positioning, management quality, brand value
- Qualitative factors may influence stock price
- Industry-specific variables not captured
Causation vs Correlation
- Regression shows association, not causation
- Cannot definitively claim improving EPS will cause stock price increase
- Recommendation: Use findings for hypothesis generation, validate with additional analysis

Future Research Directions:

Extended Analysis
- Incorporate quarterly data for more observations
- Add lagged variables to test time-delayed effects
- Include industry peer comparison analysis
Advanced Modeling
- Test non-linear relationships (polynomial regression)
- Apply time series methods (ARIMA, VAR)
- Explore machine learning models for comparison
Qualitative Factors
- Incorporate management quality metrics
- Add brand value or customer satisfaction scores
- Include ESG (Environmental, Social, Governance) metrics
Scenario Analysis
- Stress test model under different economic conditions
- Simulate impact of strategic initiatives
- Develop forecasting model for future stock price

Final Recommendations Summary:

Priority	Stakeholder	Recommendation	Expected Impact
HIGH	Management	Enhance EPS through operational efficiency	+$1.28 per 10% EPS increase
HIGH	Management	Sustain revenue growth at current trajectory	+$12.50 per $10B revenue
MEDIUM	Management	Evaluate R&D investment increase	Strengthen future growth signals
MEDIUM	Investors	Monitor quarterly EPS trends closely	Primary valuation driver
LOW	Management	Maintain current capital structure	Not a market concern currently

8. Report

Purpose

Communicate your regression analysis findings in a clear, professional, and academically rigorous format. The report demonstrates your ability to conduct statistical analysis and present actionable insights to business stakeholders.

Report Structure

Your 5-page report should follow this structure:

This is a guideline and final results will vary based on your analysis.

1. Abstract

Purpose: Provide a concise summary of the entire project for busy executives.

Content to Include:

Research Question: What business problem are you addressing?
Methodology: Brief mention of regression analysis approach
Key Findings: 2-3 most important results
Main Recommendation: Primary actionable insight

Example Abstract:

“This study examines the financial drivers of Samsung Electronics’ stock price using multiple regression analysis on 10 years of annual financial data (2014-2023). The analysis reveals that Earnings Per Share (EPS) is the strongest predictor of stock price (β = 12.80, p = 0.018), followed by revenue growth (β = 1.25, p = 0.009). The model explains 75% of stock price variation (R² = 0.75, p < 0.001). Key recommendation: Management should prioritize profitability improvement over revenue growth to maximize shareholder value. Investors should monitor quarterly EPS trends as the primary valuation signal.”

Length: 100 words (approximately 0.5 pages)

2. Introduction

Purpose: Establish business context and research motivation.

Content to Include:

Company Background: Brief overview of selected company, industry, and market position
Business Problem: Why is understanding stock price drivers important?
Research Question: State your specific research question clearly
Significance: Why should readers care about this analysis?
Report Roadmap: Preview of report sections

Example Structure:

2.1 Company Background

Samsung Electronics Co., Ltd. (KOSPI: 005930) is South Korea’s leading technology conglomerate, operating in semiconductors, consumer electronics, and mobile devices. As the world’s largest memory chip manufacturer, Samsung plays a critical role in global technology supply chains.

2.2 Research Motivation

Understanding which financial metrics drive Samsung’s stock price is crucial for: (1) investors making buy/sell decisions, (2) management prioritizing strategic initiatives, and (3) analysts forecasting future performance.

2.3 Research Question

What financial metrics most strongly predict changes in Samsung Electronics’ annual stock price?

3. Data and Methods

Purpose: Describe data sources, variables, and preprocessing steps.

Content to Include:

Data Sources

Company financial statements source (e.g., Samsung Investor Relations, 10-K filings)
Stock price data source
Time period covered (e.g., 2014-2023, 10 observations)
Data collection date

Variable Definitions

Create a table documenting all variables:

Variable Type	Variable Name	Definition	Unit	Source
Dependent	Stock_Price	Year-end closing price, adjusted for splits	USD	Yahoo Finance
Independent	Revenue	Total annual revenue	Billions USD	10-K Filing
Independent	EPS	Earnings per share (diluted)	USD	10-K Filing
Independent	Debt_to_Equity	Total debt / Total equity	Ratio	Calculated
Independent	RD_Expense	Research & development spending	Billions USD	10-K Filing
Control	Market_Return	KOSPI annual return	Percent	Korea Exchange

Data Preprocessing

Document cleaning steps:

Missing Data: “One missing EPS value for 2018 was imputed using linear interpolation”
Outliers: “2020 Net Income identified as outlier (z = 3.2) but retained due to verified one-time restructuring gain”
Transformations: “Revenue log-transformed to address right skew (skewness = 2.1)”
Adjustments: “Stock prices adjusted for 2-for-1 split in 2018”

Descriptive Statistics

Include summary statistics table (from Section 4 preprocessing output).

Length: Approximately 1 page

Purpose: Explain your analytical approach.

Content to Include:

Statistical Model

Present your regression equation:

\[\text{Stock Price}_t = \beta_0 + \beta_1(\text{Revenue})_t + \beta_2(\text{EPS})_t + \beta_3(\text{Debt/Equity})_t + \beta_4(\text{R\&D Expense})_t + \beta_5(\text{Market Return})_t + \epsilon_t\]

Variable Selection Rationale

Explain why you chose these variables:

Revenue: Measures top-line growth and market expansion
EPS: Captures profitability and earnings quality
Debt-to-Equity: Reflects financial leverage and risk
R&D Expense: Indicates innovation investment (critical for tech companies)
Market Return: Controls for overall market conditions (systematic risk)

Estimation Technique

Method: Ordinary Least Squares (OLS) regression
Software: Python (statsmodels), R (lm), or Excel (Data Analysis Toolpak)
Significance Level: α = 0.05 (95% confidence)

Assumption Testing

Linearity (residual plots)
Normality (Shapiro-Wilk test, Q-Q plots)
Homoscedasticity (Breusch-Pagan test)
Multicollinearity (VIF < 5)
Independence (Durbin-Watson statistic)

Length: Approximately 0.5 pages

4. Results

Purpose: Present statistical findings objectively.

Content to Include:

4.1 Regression Results

Present the main results table (from Section 5):

Variable	Coefficient (β)	Std Error	t-value	p-value	95% CI	Significance
(Intercept)	45.30	12.50	3.62	0.008	[18.2, 72.4]	**
Revenue (B)	1.25	0.35	3.57	0.009	[0.48, 2.02]	**
EPS	12.80	4.20	3.05	0.018	[3.45, 22.15]	*
Debt_to_Equity	-8.50	6.30	-1.35	0.219	[-22.1, 5.1]	NS
R&D_Expense (B)	3.40	1.80	1.89	0.098	[-0.75, 7.55]	†
Market_Return	2.15	0.85	2.53	0.036	[0.25, 4.05]	*

*Note: ** p<0.01, * p<0.05, † p<0.10, NS not significant*

4.2 Model Fit

Multiple R² = 0.752
Adjusted R² = 0.720
F-statistic = 23.5 (p = 0.003)
RMSE = 8.42
n = 10 observations

4.3 Assumption Test Results

Present assumption test summary table (from Section 6):

Assumption	Test	Result	Status
Linearity	Residual Plot	No pattern	✓ Met
Normality	Shapiro-Wilk	p = 0.18	✓ Met
Homoscedasticity	Breusch-Pagan	p = 0.42	✓ Met
Independence	Durbin-Watson	DW = 1.85	✓ Met
Multicollinearity	VIF	All < 5	✓ Met

4.4 Key Findings

Interpret the results in plain language:

EPS is the strongest driver: Each $1 increase in EPS is associated with a $12.80 increase in stock price (p = 0.018)
Revenue growth matters: Each $1 billion revenue increase associates with $1.25 price increase (p = 0.009)
Market sensitivity confirmed: Stock moves with overall market (β = 2.15, p = 0.036)
Debt leverage not significant: Debt-to-Equity ratio shows no significant relationship (p = 0.219)

5. Conclusion

Purpose: Synthesize findings and provide actionable recommendations.

Content to Include:

5.1 Summary of Findings

Restate key results concisely (2-3 sentences).

5.2 Business Implications

For Management: Prioritize EPS improvement through operational efficiency and margin enhancement
For Investors: Monitor quarterly EPS trends as primary valuation signal; revenue growth is secondary
Strategic Priority: Profitability over growth

5.3 Recommendations

High Priority: Implement cost optimization initiatives to boost EPS by 10% (expected stock price impact: +$1.28)
Medium Priority: Maintain revenue growth trajectory through market expansion
Low Priority: Current capital structure is adequate; debt management is not a concern

5.4 Limitations

Small sample size (n = 10) limits statistical power
Annual data masks quarterly volatility
Model doesn’t capture qualitative factors (management quality, competitive position)
Correlation ≠ causation; findings should inform, not dictate, strategy

5.5 Future Research

Extend analysis with quarterly data for more observations
Include industry peer comparison
Test non-linear relationships or machine learning models
Incorporate ESG metrics

6. Appendix

Purpose: Provide transparency and reproducibility.

Content to Include:

A. Source Code

Include complete, well-commented code for reproducibility:

#
# James, 2025/09/24
# File: Samsung_Regression.py
# Short description of the task
#

# 1. Input
# Import libraries
import pandas as pd
import numpy as np
import statsmodels.api as sm
from scipy import stats

# Load data
df = pd.read_csv("samsung_financial_data.csv")

# Data preprocessing
# [Include all preprocessing steps with comments]

# Regression analysis
X = df[['Revenue', 'EPS', 'Debt_to_Equity', 'RD_Expense', 'Market_Return']]
y = df['Stock_Price']

# 2. Process
X = sm.add_constant(X)
model = sm.OLS(y, X).fit()

# 3. Output
# Display results
print(model.summary())

# Assumption testing
# [Include diagnostic tests]

# Visualization
# [Include plots]

B. Additional Tables (if needed)

Correlation matrix
Full VIF table
Robustness checks

C. Data Sources Reference

Full URLs and access dates for all data sources
Specific filing references (e.g., “Samsung Electronics 2023 Annual Report, page 45”)

Length: Code appendix doesn’t count toward 5-page limit

Formatting Guidelines

Citations:

Cite data sources and methodological references

9. Presentation

Format

Duration: 3-minute presentation + 1-minute Q&A (4 minutes total)
Style: Verbal presentation only (TED talk format)
Visual Aids: No slides permitted

Key Points

Focus on clear verbal delivery and storytelling
Engage the audience through your voice, body language, and presence
Structure your content with a strong opening, clear main points, and memorable conclusion
Practice timing to ensure you stay within the 3-minute limit

10. Submission

Submit both documents as PDF files through iLearn:
- Report (6 pages, PDF format)
- Presentation slides (15 slides, PDF format)

Example Project Ideas

“Regression Analysis of Earnings and Stock Price for Samsung Electronics”
“How Debt Influences Tata Motors’ Stock Performance”
“R&D Expense as a Predictor of Novo Nordisk’s Market Value”

Project Title

Objective

Project Workflow

1. Company & Question Selection

Step-by-Step Instructions: Selecting a Publicly Listed Company

1. Brainstorm and Shortlist

2. Confirm Public Listing Status

3. Check Data Availability

4. Create a Short Rationale

Example (for Samsung Electronics, Korea)

Research Question

2. Data Acquisition

Purpose

Metrics

Expected Output

3. Dataset Construction

Purpose

Metrics

Expected Output

4. Data Verification & Preprocessing

Purpose

Metrics

Expected Output

5. Regression Analysis

Purpose

Metrics

Expected Output

6. Model Evaluation

Purpose

Metrics

Expected Output

7. Insights & Recommendations

Purpose

Metrics

Expected Output

8. Report

Purpose

Report Structure

1. Abstract

2. Introduction

3. Data and Methods

4. Results

5. Conclusion

6. Appendix

Formatting Guidelines

Submission Requirements

9. Presentation

Format

Key Points

10. Submission

Example Project Ideas