Business Research Methods

Module-IV
Module 4: Hypothesis Testing
Review of Data Presentation – Graphs, Charts and Tables*,
Testing of Hypothesis-Z Test; t-test (Independent and
Paired samples),
F-Test: One-way and Two-way ANOVA, Chi-Square Test: Non -Parametric tests - Categorical
Data
(Independence of Attributes),
Quantitative Data (Goodness of Fit),
Technology usage for Data Analysis and
Hypothesis Testing, Review of report writing*.
Data Representation

 From a research and academic perspective, data representation techniques (or

tools for presenting data) are essential to make raw data interpretable and
meaningful.

 They can be broadly classified into tabular, graphical, and diagrammatic methods.
 1. Tabular Representation
Data is organized systematically in rows and columns to facilitate easy understanding.

 Simple Tables: List raw data for each observation (e.g., monthly stock returns).
 Frequency Distribution Tables: Show how often each value or range of values occurs.
 Example: Number of companies falling in different market capitalization ranges.
 Cumulative Frequency Tables: Show accumulation of frequencies up to a certain
point.
 Cross-tabulation / Contingency Tables: Show relationships between two or more
categorical variables.
 Advantages: Precise, easy to record, and useful for small or large datasets.
2. Graphical Representation
Graphical methods help visualize patterns, trends, and relationships.
Common Graphical Tools:
Bar Chart: Compares quantities across categories (vertical or horizontal).
Histogram: Shows frequency distribution of numerical data; bars touch each other to
indicate continuous data.
Pie Chart: Represents proportions of a whole; each slice corresponds to a category.
Line Graph: Shows trends over time; commonly used for stock prices, GDP growth, etc.
Dot Plot / Scatter Plot: Shows relationship between two variables; useful for correlation
analysis.
Stem-and-Leaf Plot: Shows distribution while retaining actual data values.
Area Chart: Similar to line chart, emphasizes magnitude of values over time.
3. Diagrammatic or Pictorial Representation
Flowcharts: Represent processes or sequences of events.

Maps / Spatial Diagrams: Show geographical distribution (e.g., economic activity by

region).

Infographics / Pictograms: Use symbols or pictures to represent quantities for better

visualization.
 4. Advanced / Statistical Representation Techniques
For research purposes, especially in economics, finance, or social sciences:

Box-and-Whisker Plot (Boxplot): Visualizes median, quartiles, outliers, and spread.

Control Charts: Used in quality control; shows variation over time.

Time-Series Plots: Shows trends, cycles, and seasonal effects.

Heat maps / Correlation Matrices: Visualize relationships among multiple variables

simultaneously.

Histograms with Density Curves: Combine frequency and probability distribution.

 Key Principles for Effective Data Presentation
•Clarity: Avoid clutter; highlight key points.
•Accuracy: Ensure numbers and scales are correct.
•Relevance: Choose representation suited to your data type and research
objective.
•Comparability: Use consistent scales, especially for multiple datasets.
•Interpretability: Graphs and tables should allow the reader to draw
conclusions easily.
Hypothesis:
A hypothesis is a tentative statement or proposition that predicts a relationship
between two or more variables, which can be tested empirically through observation or
experimentation.

It is essentially a proposed explanation or assumption that guides the research study.

Key Features of a Hypothesis

Testable: It should be possible to verify or falsify the hypothesis using data.

Specific: Clearly defines the variables and the expected relationship between them.

Predictive: Suggests an outcome or effect that the study seeks to confirm.

Based on Theory or Observation: It is grounded in prior research, literature, or logical

reasoning.
Types of Hypotheses
Null Hypothesis (H₀):
States that there is no relationship or effect between variables.
Example: “The Hindenburg report had no impact on Adani stock prices.”
Alternative Hypothesis (H₁ or Ha):
States that there is a relationship or effect.
Example: “The Hindenburg report caused a significant decline in Adani stock prices.”
Directional Hypothesis:
Specifies the direction of the relationship.
Example: “The stock price of Adani Enterprises declined after the Hindenburg report.”
Non-Directional Hypothesis:
Predicts a relationship without specifying direction.
Example: “The Hindenburg report affected Adani stock prices.”
 Important insights and nuances about hypotheses
1. A Hypothesis is Not a Guess
Many beginners think a hypothesis is a random assumption.
In reality, it’s an informed, logical proposition based on theory, prior studies, or
observation.
Example: Using EMH theory to hypothesize about abnormal stock returns after the
Hindenburg report.
2. Hypotheses Guide the Entire Research
They define what data to collect, how to analyze it, and what statistical tests to apply.
A poorly framed hypothesis leads to vague or inconclusive results.
3. Null vs Alternative Hypothesis
The null hypothesis (H₀) is the default assumption: “nothing happens” or “no relationship
exists.”
The alternative hypothesis (H₁) reflects the expected effect or relationship.
Statistical testing is essentially a decision on whether to reject H₀ in favor of H₁.
4. Directional vs Non-Directional
Directional hypothesis: Predicts the direction of the effect (increase/decrease).
Non-directional hypothesis: Only predicts a change, not the direction.
Directional hypotheses can increase statistical power but must be justified theoretically.
5. Hypotheses Can Be Quantitative or Qualitative
Quantitative: Involves measurable variables (e.g., stock returns, revenue).
Qualitative: Involves concepts or categories (e.g., investor sentiment, corporate governance
quality).
6. A Good Hypothesis is Testable and Falsifiable
Testability: You should be able to collect data to verify it.
Falsifiability: There should be a possibility of proving it wrong.
This is critical for scientific rigor.
7. Hypotheses Evolve
Sometimes, initial hypotheses may be refined after exploratory data analysis or
pilot studies.
This is normal and part of the research process, as long as changes are
documented and justified.
 Definition of Z-Test
1. Large Sample Size (n > 30)
•Rule of thumb: Z-test is appropriate when the sample size is greater than 30.
•Reason: By the Central Limit Theorem, the sampling distribution of the mean is approximately normal, even if
the population is not perfectly normal.

2. Small Sample Size (n ≤ 30)

•If the population standard deviation (σ) is known, a Z-test can still be used.
•If σ is unknown, it’s better to use a t-test, which adjusts for the small sample uncertainty.

3. Other Considerations
•For proportion tests, the rule is slightly different:
•np ≥ 5 and n(1-p) ≥ 5, where p is the population proportion.
•The key idea: Z-test relies on the normal approximation of the sampling distribution.
 Scenario:
A firm claims that the average accounts receivable turnover ratio for its
industry peers is 8. You want to test whether your company’s turnover ratio
differs from this industry average.
Data:
Sample size: 50 companies (n > 30 → large sample)
Sample mean turnover ratio: 7.5
Population standard deviation (σ): 1.2
Hypotheses:
H₀: µ = 8 (no difference from industry average)
H₁: µ ≠ 8 (different from industry average)

Interpretation:
|Z| > 1.96 → Reject H₀ → The company’s turnover ratio is significantly different from the industry
average.
 Population
Test Purpose Sample Size / Conditions Notes / Examples
Variance Known?
Test if a sample mean differs from One-sample mean, two-sample mean,
σ known (or large
Z-Test population mean or compare two Large sample (n > 30) proportion tests; e.g., abnormal returns
sample)
means in stock event study

Test if a sample mean differs from One-sample t-test, independent two-

t-Test (Student’s t) population mean or compare two Small sample (n ≤ 30) σ unknown sample t-test, paired t-test; e.g., testing
means mean returns of a small stock sample

Compare means of related samples Any sample size (preferably Example: Pre- and post-event stock
Paired t-Test σ unknown
(before/after) small) returns

Example: Testing homogeneity of

Compare variances of two or more
F-Test Any sample size σ unknown variance for returns across different
populations
stocks; also used in ANOVA

Test relationships between

Example: Frequency of bullish vs
Chi-Square (χ²) Test categorical variables or goodness-of- Any sample size σ not needed
bearish days in a month
fit
ANOVA (F-test in Compare means of more than two Example: Comparing average returns of
Any sample size σ unknown
ANOVA) groups multiple sectors
Test difference between population Large sample (np ≥5, n(1-p) σ can be derived Example: % of days a stock closed
Proportion Z-Test
proportions ≥5) from proportion above moving average
 Category Software / Tools Purpose / Features

Widely used for statistical tests,

Quantitative Data Analysis SPSS
regression, ANOVA, survey data analysis.

Strong in econometrics, time-series, panel

STATA
data analysis.

Specialized for econometric and time-

EViews
series modeling.
Open-source, advanced statistical
R
computing and visualization.

Flexible for statistics, machine learning,

Python econometrics (NumPy, pandas,
statsmodels).

Powerful for advanced statistical modeling

SAS
and large datasets.
High-level computing, simulation,
MATLAB
mathematical modeling.
Basic statistical analysis, data visualization,
Excel
pivot tables.
Steps in Hypothesis Testing
Step Explanation

- Null hypothesis (H₀): No effect or no difference (e.g.,

“There is no relationship between FDI and GDP”).
1. State the hypotheses
- Alternative hypothesis (H₁): There is an effect or
difference (e.g., “FDI significantly influences GDP”).

Commonly 0.05 (5%) → means you allow a 5% risk of

2. Choose the significance level (α)
rejecting H₀ when it is true.

Based on type of data and research design:

- t-test (compare means of two groups)
- ANOVA (compare means across more than two
groups)
3. Select the appropriate test
- Chi-square test (categorical data, association)
- Correlation/Regression (relationship between
variables)
- Z-test (large samples, population parameters known)