Unit 9: Hypothesis Testing - Subjective Questions

Definition: A hypothesis is a tentative, logical, and testable statement about the expected relationship between two or more variables. It is formulated on the basis of theory, previous research, or observation.

Characteristics of a good hypothesis:

• It should be clear, specific, and stated in simple language.
• It must be empirically testable through the collection and analysis of data.
• It should identify the variables involved and indicate their expected relationship.
• It must be consistent with existing theories and known facts.
• It should be capable of being accepted or rejected through statistical testing.
• It should be limited in scope and related to the research problem.

Example: There is a significant difference in the average examination scores of students who receive online instruction and those who receive traditional classroom instruction.

The hypothesis testing process generally includes the following steps:

• State the hypotheses: Formulate the null hypothesis and the alternative hypothesis or .
• Select the level of significance: Choose a value such as or .
• Select the appropriate test statistic: Choose a -test, -test, or another test based on the sample size, population variance, and research design.
• Establish the decision rule: Determine the critical value or rejection region.
• Collect and analyze data: Calculate the required sample statistics and test statistic.
• Make a statistical decision: Reject if the test statistic falls in the rejection region or if the -value is less than .
• State the conclusion: Interpret the result in relation to the research question and explain whether sufficient evidence exists to support .

Null hypothesis (): The null hypothesis states that there is no significant difference, relationship, or effect in the population. It usually contains an equality sign.

Alternative hypothesis (): The alternative hypothesis states that a significant difference, relationship, or effect exists. It represents the claim that the researcher wants to investigate.

Example: Suppose a researcher wants to determine whether a training program increases employee productivity.

• , meaning the training program has no effect.
• , meaning the training program increases productivity.

The statistical test evaluates the evidence against . Researchers generally do not prove ; they either reject it or fail to reject it.

Important types of hypotheses include:

• Null hypothesis: States that no significant relationship or difference exists between variables.
• Alternative hypothesis: States that a significant relationship or difference exists.
• Directional hypothesis: Predicts the direction of the relationship or difference, such as an increase or decrease.
• Non-directional hypothesis: Predicts that a difference exists but does not specify its direction.
• Simple hypothesis: Involves one independent variable and one dependent variable.
• Complex hypothesis: Involves two or more independent variables, dependent variables, or both.
• Associative hypothesis: States that variables change together without claiming that one causes the other.
• Causal hypothesis: Predicts that a change in one variable produces a change in another variable.
• Statistical hypothesis: Expresses a research claim in terms of population parameters such as , , or .

The type selected depends on the research problem, variables, and study design.

• Level of significance: The level of significance, denoted by , is the maximum probability of rejecting a true null hypothesis. Common values are and .
• Critical region: This is the set of values of the test statistic for which the null hypothesis is rejected.
• Critical value: It is the boundary value that separates the acceptance or non-rejection region from the rejection region.
• p-value: The p-value is the probability of obtaining a result at least as extreme as the observed result, assuming that is true.

Decision rule:

• If , reject .
• If , fail to reject .

A smaller p-value indicates stronger evidence against the null hypothesis.

A one-sample -test is used to determine whether the mean of a sample differs significantly from a specified population mean when the population standard deviation is unknown.

Procedure:

• State the hypotheses, for example and .
• Select the significance level .
• Calculate the sample mean and sample standard deviation .
• Compute the test statistic:
  
• Calculate the degrees of freedom: .
• Obtain the critical value from the distribution table or calculate the p-value.
• Reject if the calculated value lies in the rejection region or if .
• State the conclusion in the context of the research problem.

The one-sample -test compares the observed sample mean with a hypothesized population mean .

The difference between the sample mean and hypothesized mean is standardized by the estimated standard error:

where:

• is the sample mean.
• is the hypothesized population mean.
• is the sample standard deviation.
• is the sample size.
• is the estimated standard error of the mean.

The degrees of freedom are:

Assumptions:

• The observations are independent.
• The dependent variable is measured at least on an interval scale.
• The population is approximately normally distributed, especially when the sample is small.
• The population standard deviation is unknown and is estimated using .

An independent two-sample -test is used to compare the means of two unrelated populations.

Procedure:

• State the hypotheses, such as and .
• Select the significance level .
• Obtain the sample means and , sample variances and , and sample sizes and .
• If equal variances are assumed, calculate the pooled variance:
  
• Calculate the test statistic:
  
• Use .
• Compare the calculated statistic with the critical value or use the p-value.
• Draw a conclusion about the difference between the two population means.

A paired-samples -test is used when two observations are related, such as measurements taken from the same participants before and after an intervention.

Procedure:

• Calculate the difference for each pair: .
• Calculate the mean difference and standard deviation of differences .
• State the hypotheses, such as and .
• Calculate the test statistic:
  
• Use .
• Compare the test statistic with the critical value or evaluate the p-value.

Difference from an independent-samples test:

• A paired test analyzes within-pair differences.
• An independent test compares the means of two unrelated groups.
• A paired test accounts for the dependence between observations and often reduces error caused by individual differences.

A one-sample -test is used to test a claim about a population mean when the population standard deviation is known or when the sample is sufficiently large.

Procedure:

• State the hypotheses, for example and .
• Select the significance level .
• Calculate the sample mean .
• Use the known population standard deviation and sample size .
• Compute the test statistic:
  
• Identify the critical value for a one-tailed or two-tailed test.
• Reject if the calculated value lies in the rejection region or if .
• Interpret the result in relation to the population mean.

A two-sample -test is used to examine whether the means of two populations differ significantly when the population standard deviations are known or the samples are sufficiently large.

Procedure:

• State the hypotheses, such as and .
• Select the significance level .
• Calculate the sample means and .
• Determine the standard error:
  
• Calculate the test statistic:
  
  where is the hypothesized difference, usually .
• Compare the calculated value with the critical value or use the p-value.
• Make and interpret the statistical decision.

One-tailed test:

• It is used when the alternative hypothesis specifies a direction.
• Examples are or .
• The entire rejection region of size is placed in one tail of the distribution.
• It is appropriate when only an increase or only a decrease is relevant.

Two-tailed test:

• It is used when the alternative hypothesis states that a difference exists but does not specify its direction.
• The alternative hypothesis is .
• The rejection region is divided between both tails, with in each tail.
• It is appropriate when both increases and decreases are important.

The choice must be made before examining the data because changing the tail after seeing the results can produce misleading conclusions.

Type I error: A Type I error occurs when the null hypothesis is rejected even though it is true. Its probability is denoted by .

Type II error: A Type II error occurs when the null hypothesis is not rejected even though it is false. Its probability is denoted by .

Statistical power: The probability of correctly rejecting a false null hypothesis is called power:

Factors influencing errors:

• Increasing generally increases the chance of rejecting and may reduce .
• Increasing sample size generally reduces both sampling uncertainty and .
• A larger true difference between population parameters increases power.
• Greater population variability makes significant differences more difficult to detect.
• A one-tailed test may have greater power in the specified direction than a two-tailed test.

Assumptions common to both tests:

• The observations should be randomly selected or reasonably representative.
• Observations should be independent unless a paired design is specifically being used.
• The variables should be measured quantitatively.

Additional conditions for t-tests:

• The population standard deviation is generally unknown.
• For small samples, the population should be approximately normal.
• An independent two-sample t-test may require equal population variances when the pooled version is used.
• A paired t-test requires meaningful pairing and approximately normally distributed differences.

Additional conditions for z-tests:

• The population standard deviation should be known for an exact one-sample z-test.
• Large samples may justify the use of the normal approximation through the central limit theorem.
• For a two-sample z-test, the standard errors or population variances must be known or appropriately estimated under large-sample conditions.

Violating assumptions can affect the reliability of p-values and conclusions.

A two-proportion z-test determines whether two population proportions differ significantly.

Let the sample proportions be and .

Procedure:

• State the hypotheses, such as and .
• Calculate the pooled proportion under the null hypothesis:
  
• Calculate the standard error under :
  
• Calculate the test statistic:
  
• Compare the statistic with the critical value or use the p-value.
• The normal approximation is generally appropriate when the expected successes and failures in each group are sufficiently large.

The three situations differ in the structure of the data and the appropriate test statistic.

• Single population: One sample mean is compared with a hypothesized population mean. A one-sample -test or -test is used. The parameter of interest is .
• Two independent populations: Means from two unrelated groups are compared. An independent two-sample -test or two-sample -test is used. The parameter of interest is .
• Paired samples: Two observations are collected from the same unit or from matched units. The analysis is based on the differences within each pair, and a paired-samples -test is commonly used. The parameter of interest is .

Using an independent test for paired observations or ignoring dependence can produce an incorrect standard error and misleading conclusions.

A one-sample proportion test is used to determine whether a population proportion differs from a specified value.

Procedure:

• State the hypotheses, such as and .
• Calculate the sample proportion:
  
• Select the significance level .
• Calculate the standard error under the null hypothesis:
  
• Calculate the test statistic:
  
• Compare the calculated value with the relevant critical value or determine the p-value.
• Reject when the evidence is statistically significant.
• State the conclusion in terms of the population proportion.

The normal approximation should be used only when the expected numbers of successes and failures are adequate.

Statistical significance indicates that the observed result would be unlikely if the null hypothesis were true. It is commonly determined by comparing the p-value with .

Practical significance concerns the real-world importance or usefulness of the observed difference or relationship. A result can be statistically significant but too small to matter in practice.

Reasons to consider both:

• A very large sample can make a trivial difference statistically significant.
• A small sample may fail to produce statistical significance even when a meaningful effect exists.
• Practical significance helps decision-makers assess whether an intervention, policy, or relationship is important enough to act upon.
• Effect size and confidence intervals provide useful information about the magnitude and precision of a result.

Therefore, researchers should report the p-value together with effect size, confidence intervals, and substantive interpretation.

Confidence intervals and hypothesis tests are closely related methods for making inferences about population parameters.

• A confidence interval provides a range of plausible values for a population parameter.
• A hypothesis test evaluates whether a specified parameter value is consistent with the sample data.
• For a two-tailed test at significance level , a corresponding confidence interval can be used.
• If the hypothesized value lies outside the confidence interval, is rejected at the significance level.
• If the hypothesized value lies inside the interval, is not rejected at that level.

For example, if a confidence interval for is , it does not contain . This supports rejection of in a two-tailed test at .

Step 1: State the hypotheses

Step 2: Select the test

The population standard deviation is unknown and the sample size is small, so a one-sample -test is appropriate.

Step 3: Calculate the test statistic

The degrees of freedom are .

Step 4: Determine the critical value

For a two-tailed test with and degrees of freedom, the critical values are approximately .

Step 5: Decision and conclusion

Since , fail to reject . There is insufficient evidence at the level to conclude that the population mean score differs from .