Hypothesis Testing Notes

Hypothesis Testing

Introduction to Hypothesis

In ordinary terms, a hypothesis is an assumption. However, for a researcher, it's a formal question intended to be resolved. It is defined as a proposition set forth as an explanation for the occurrence of some specified group of phenomena. A research hypothesis is often a predictive statement that can be tested using scientific methods, relating an independent variable to a dependent variable.

  • Examples:

    • "Students who receive counselling will show a greater increase in creativity than students not receiving counselling."

    • "The automobile A is performing as well as automobile B."

A hypothesis states what we are looking for and can be tested to determine its validity.

Characteristics of a Good Hypothesis

  1. Clarity and Precision: A hypothesis should be clear and precise to ensure reliable inferences.

  2. Testability: It must be capable of being tested. Prior studies might be necessary to make a hypothesis testable. A hypothesis is testable if deductions can be made from it, which can be confirmed or disproved by observation.

  3. Relationship Between Variables: A relational hypothesis should state the relationship between variables.

  4. Limited Scope and Specificity: Narrower hypotheses are generally more testable.

  5. Simplicity: It should be stated in simple terms for easy understanding, though simplicity does not dictate significance.

  6. Consistency with Known Facts: It should be consistent with established facts and accepted by experts.

  7. Amenability to Testing Within a Reasonable Time: Excellent hypotheses should be testable within a reasonable time frame.

  8. Explanatory Power: It should explain the facts that gave rise to the need for explanation and have empirical reference.

Basic Concepts in Hypothesis Testing

Null Hypothesis and Alternative Hypothesis

  • Null Hypothesis (H₀): An assumption that there is no significant difference between populations or variables.

  • Alternative Hypothesis (Hₐ): A statement that contradicts the null hypothesis, suggesting a significant difference or relationship.

If comparing method A with method B, the null hypothesis would be that both methods are equally good. The alternative hypothesis could be that method A is superior or method B is inferior.

  • Example: Testing if the population mean μ\mu is equal to a hypothesized mean μH0=100\mu_{H0} = 100.

    • Null Hypothesis: H<em>0:μ=μ</em>H0=100H<em>0: \mu = \mu</em>{H0} = 100

    • Alternative Hypothesis:

      • Ha:μ≠100H_a: \mu \neq 100 (Two-tailed)

      • H_a: \mu > 100 (Right-tailed)

      • H_a: \mu < 100 (Left-tailed)

The null and alternative hypotheses are chosen before the sample is drawn to avoid deriving hypotheses from the data.

Considerations for Choosing the Null Hypothesis:

  1. The alternative hypothesis is usually the one a researcher wishes to prove, while the null hypothesis is the one to disprove.

  2. If rejecting a true hypothesis involves great risk, it should be the null hypothesis. The probability of rejecting it when true is α\alpha (level of significance), chosen to be very small.

  3. The null hypothesis should always be specific.

Types of Hypotheses

  1. Null Hypothesis: As explained above.

  2. Alternative Hypothesis: As explained above.

  3. Simple & Composite Hypothesis:

    • Simple (Specific) Hypothesis: Specifies an exact value. Example: H0:μ=10H_0: \mu = 10. The Null hypothesis specifies that the mean is exactly 10.

    • Composite (Non-Specific) Hypothesis: Does not specify an exact value. Example: H0:μ≠10H_0: \mu \neq 10. The Null hypothesis specifies that the mean is not 10, allowing for a range of values.

  4. Directional and Nondirectional Hypotheses:

    • Directional Hypotheses: State the direction of the relationship between variables (positive/negative) or the nature of the difference between groups (more than/less than).

      • Example:

        • “The greater the stress experienced in the job, the lower the job satisfaction of employees.”

        • “Women are more motivated than men.”

    • Nondirectional Hypotheses: Postulate a relationship or difference but do not indicate the direction.

      • Example:

        • “There is a relation between arousal‐seeking tendency and consumer preferences for complex product designs.”

        • “There is a difference between the work ethic values of American and Asian employees.”

Nondirectional hypotheses are used when relationships or differences have not been explored, there is no basis for indicating the direction, or there are conflicting findings in previous research.

Other Basic Concepts

  1. Level of Significance ($\alpha$): The probability of rejecting the null hypothesis when it is true. It is usually set at 5% (0.05), meaning there is a 5% risk of rejecting H₀ when it is true.

  2. Confidence Interval: The range within which the true population parameter is expected to fall. If the level of significance is 5%, the confidence interval is 95%. This is conventionally accepted level for most business research, expressed as p≤0.05p \le 0.05.

  3. Decision Rule or Test of Hypothesis: A rule to decide whether to accept H₀ (reject Hₐ) or reject H₀ (accept Hₐ). For example, testing 10 items and accepting H₀ if there are none or only 1 defective item.

  4. Type I and Type II Errors:

    • Type I Error ($\alpha$ error): Rejecting H₀ when it is true (false positive).

    • Type II Error ($\beta$ error): Accepting H₀ when it is false (false negative).

Decision

H₀ is True

H₀ is False

Accept H₀

Correct

Type II Error ($\beta$)

Reject H₀

Type I Error ($\alpha$)

Correct

Controlling Type I error involves setting a lower level, but reducing Type I error increases the probability of Type II error. Decision-makers balance these errors based on their associated costs.

Procedure for Hypothesis Testing

The process involves making a formal statement, selecting a significance level, deciding on the distribution to use, selecting a random sample, calculating the probability, and comparing the probability with the significance level.

  1. Making a Formal Statement: State the null hypothesis (H₀) and the alternative hypothesis (Hₐ) clearly.

    • Example:

      • Null hypothesis H0:μ=10H_0 : \mu = 10 tons

      • Alternative Hypothesis H_a : \mu > 10 tons

  2. Selecting a Significance Level: Choose a pre-determined level of significance (e.g., 5% or 1%).

  3. Deciding the Distribution to Use: Determine the appropriate sampling distribution (normal distribution or t-distribution).

  4. Selecting a Random Sample and Computing an Appropriate Value: Select a random sample and compute the test statistic.

  5. Calculation of the Probability: Compute the probability that the sample result diverges as widely as it has from expectations, assuming the null hypothesis is true.

  6. Comparing the Probability: Compare the calculated probability with the specified value for α\alpha, the significance level. If the calculated probability is equal to or smaller than α\alpha (for a one-tailed test) or α/2\alpha/2 (for a two-tailed test), reject the null hypothesis. Otherwise, accept the null hypothesis.

Tests of Hypotheses

Tests of hypotheses, also known as tests of significance, help determine the validity of an assumption about a population parameter.

Types of Tests

  1. Parametric Tests: Assume certain properties of the parent population, such as observations coming from a normal population, large sample size, and assumptions about population parameters.

  2. Non-Parametric Tests: Do not depend on assumptions about the parameters of the parent population.

Specific Tests

  • z-test: Based on the normal probability distribution. Used for judging the significance of statistical measures, especially the mean.

  • t-test: Based on t-distribution. Appropriate for judging the significance of a sample mean or the difference between means of two samples when the sample size is small and the population variance is unknown.

Hypothesis Testing of Means

The testing technique differs based on various situations such as population normality, sample size, and knowledge of population variance.

  1. Population Normal, Population Infinite, Variance Known: Use z-test.

When sample size is less than 30 we apply t-test

Hypothesis Testing for Differences Between Means

Useful for determining whether the parameters of two populations are alike or different. The null hypothesis is generally stated as H<em>0:μ</em>1=μ<em>2H<em>0: \mu</em>1 = \mu<em>2, where μ</em>1\mu</em>1 and μ2\mu_2 are the population means of the two populations.

  1. Population Variances Known or Large Samples: Use z-test.

  2. Small Samples with Unknown Population Variances: Use t-test when sample size is less than 30.

Chi-Square Test

The chi-square test (χ2\chi^2) is a statistical measure used in sampling analysis for comparing a variance to a theoretical variance. It is a non-parametric test used to determine if categorical data shows dependency or if two classifications are independent. It is also used for comparisons between theoretical populations and actual data when categories are used.

Applications:

  • Test the goodness of fit.

  • Test the significance of association between two attributes.

  • Test the homogeneity or the significance of population variance.

If the calculated value of χ2\chi^2 is less than the table value at a certain level of significance, the fit is considered good. If it is greater, the fit is not considered good.

Conditions for Application:

  1. Observations recorded and used are collected on a random basis.

  2. All items in the sample must be independent.

  3. No group should contain very few items (less than 5). Regrouping is done if frequencies are less than 5.

  4. The overall number of items must be reasonably large (at least 50).

  5. The constraints must be linear.

Important Characteristics:

  • Based on frequencies, not parameters.

  • Used for testing hypotheses, not estimation.

  • Possesses the additive property.

  • Can be applied to complex contingency tables.

  • Important non-parametric test with no rigid assumptions.

Analysis of Variance (ANOVA)

ANOVA is used when multiple sample cases are involved to examine the significance of the difference amongst more than two sample means simultaneously. It is used to draw inferences about whether samples have been drawn from populations having the same mean.

Basic Principle

The basic principle of ANOVA is to test for differences among the means of the populations by examining the amount of variation within each of these samples, relative to the amount of variation between the samples.

ANOVA Technique: One-way

Under the one-way ANOVA, only one factor is considered. The technique involves the following steps:

Short-Cut Method for One-Way ANOVA

ANOVA can be performed using a short-cut method, which is practical and convenient. The various steps involved in the shortcut method are as under:

This ratio is used to judge whether the difference among several sample means is significant or is just a matter of sampling fluctuations. For this purpose, we investigate the table, giving the values of F for given degrees of freedom at different levels of significance. If the worked- out value of F is less than the table value of F, the difference is taken as insignificant i.e., due to chance and the null-hypothesis of no difference between sample means stands. In case the calculated value of F happens to be either equal or more than its table value, the difference is considered as significant (which means the samples could not have come from
the same universe) and accordingly the conclusion may be drawn. The higher the calculated value of F is above the table value, the more definite and surer one can be about his conclusions.

V = varieties
n = plots
Illustration 1
Set up an analysis of variance table for the following per acre production data for
three varieties of wheat, each grown on 4 plots and state if the variety differences are
significant.

Introduction to data analysis using Excel

What is Data Analysis?

Data analysis is the process of inspecting, cleaning, transforming, and modeling data to discover useful information, draw conclusions, and support decision-making.

Why Use Excel for Data Analysis?

Excel is a widely used tool because it’s:
User-friendly
Easily accessible
Packed with built-in functions
Great for basic and intermediate analysis

Key Components of Excel for Data Analysis

Data Entry and Cleaning

Entering Data: Use rows (records) and columns (fields/variables)
Cleaning: Remove duplicates, fix spelling errors, fill blanks
Use features like:
Text to Columns
Find & Replace
Data Validation

Sorting and Filtering

Sort data A–Z or Z–A (numbers, text, or dates)

Filter to display only the rows that meet certain conditions

Basic Functions

SUM() – Adds a range of values
AVERAGE() – Finds the mean
MAX() / MIN() – Finds the highest/lowest value
COUNT() / COUNTA() – Counts numbers or non-empty cells

Using Formulas

Basic formula: =A1+B1
Use relative (A1) and absolute references ($A$1)

Conditional Formatting

Highlight cells based on rules (e.g., values above a threshold)
Great for visual analysis

Charts and Graphs

Create visuals to understand trends:
Column Chart
Line Chart
Pie Chart
Scatter Plot
Use Insert > Chart

Pivot Tables

Powerful tool to summarize data
Drag and drop fields to analyze large datasets easily
Found in Insert > PivotTable

Data Analysis Toolpak (Optional Add-in)

Go to: File > Options > Add-ins
Enable “Analysis ToolPak” for tools like:
Descriptive statistics
Regression
t-Test
ANOVA

Conclusion

Excel provides a solid foundation for basic data analysis tasks. Once you're comfortable with sorting, filtering, formulas, and charts, you can move into more advanced tools like PivotTables and the Analysis Toolpak.