Statistics 11/19 Test 4

Overview of Hypothesis Testing

Introduction to Hypothesis Testing

  • Definition of hypothesis testing: The process of making a decision regarding the validity of a hypothesis based on sample data.

  • Centrality in Statistics: Hypothesis testing is a key concept in statistics that allows making decisions based on data while accounting for uncertainty.

Importance of Confidence in Sample Data

  • Uncertainty in Statistics: No statistical statement can be said to be 100% true or false due to inherent variability.

  • Sampling: When taking a sample, it should represent the population adequately, as highlighted by sampling distributions.

  • Confidence Intervals: For example, using a 95% confidence level means we can be confident that the sample will capture the population mean or proportion in 95 out of 100 samples.

Framework of Hypothesis Testing

  • Steps in Hypothesis Testing:

    • Starting with a hypothesis question.

    • Gathering sample statistics such as mean and proportion to formulate tests around.

    • Implementing a form of confidence interval based on the sample statistics.

Confidence Intervals in Hypothesis Testing

  • If a sample mean (X̄) and standard deviation (s) are observed:

    • Confidence interval is calculated for the normal distribution:
      [Xˉ−z<em>racα2(s/extsqrt(n)),Xˉ+z</em>racα2(s/extsqrt(n))][X̄ - z<em>{ rac{α}{2}}(s/ ext{sqrt}(n)), X̄ + z</em>{ rac{α}{2}}(s/ ext{sqrt}(n))]

    • Example Values: Given a confidence interval extending to ±1.96 from the mean, we can state it captures the population mean within that range.

Interpreting Sample Statistics

  • The question arises: Does the sample mean (X̄) fall within the established confidence interval?

  • If X̄ is within the interval:

    • Population mean and sample mean are close enough for us to consider them similar, concluding that there's no significant difference.

  • If X̄ falls outside:

    • Indicates a significant difference from the population mean.

Formulating Hypotheses

  • Null Hypothesis (H₀):

    • Represents a statement of no change (the status quo).

    • Always assumes that the population parameter equals the sample statistic.

    • Example: If stated that population mean = 25, then H₀: μ = 25.

  • Alternative Hypothesis (H₁):

    • Opposes H₀, attempting to provide evidence of a change or effect.

    • May be of three forms:

    1. Less than (H₁: μ < 25) - One tail test (left side).

    2. Greater than (H₁: μ > 25) - One tail test (right side).

    3. Not equal to (H₁: μ ≠ 25) - Two-tailed test.

Type of Tests and Decision Making

  • One-Tailed Tests:

    • Focused on one side of the distribution based on hypotheses that specify the direction of the difference.

  • Two-Tailed Tests:

    • Checking if the sample mean differs significantly in both directions from the population mean.

Real-World Analogy: Hypothesis Testing as a Jury Trial

  • Jury Trial Example:

    • Null Hypothesis: The accused is not guilty (H₀).

    • Alternative Hypothesis: The accused is guilty (H₁).

    • Evidence presented can either reject or fail to reject the null hypothesis.

Possible Outcomes:
  1. If H₀ is true and we do not reject it: Correct Decision (not guilty).

  2. If H₁ is true and we reject H₀: Correct Decision (guilty).

  3. If H₁ is true but we do not reject H₀: Type II Error (not guilty when guilty).

  4. If H₀ is true but we reject it: Type I Error (guilty when not guilty).

Understanding Type I and Type II Errors

  • Type I Error (α):

    • Rejecting the null hypothesis when it is true.

    • Example: Concluding more than 2% of children experience side effects when in fact they do not.

  • Type II Error (β):

    • Failing to reject the null hypothesis when it is false.

    • Example: Concluding less than 2% have side effects when actually it is more.

Minimizing Type II Error

  • Strategizing against Errors: The aim is to keep the probability of Type II error (β) less than the level of significance (α) provided in the problem.

Practical Example of Hypothesis Testing

  • Case Study - New Antibiotic:

    • Null Hypothesis (H₀): The rate of headaches is 2% (0.02).

    • Alternative Hypothesis (H₁): The rate of headaches is greater than 2%.

  • Implications of Type Errors:

    • Type I Error: Rejecting H₀ when the antibiotic is actually safe (unnecessary rejection).

    • Type II Error: Failing to reject H₀ when more than 2% experience headaches, possibly ignoring harmful effects.

Conclusion and Next Steps

  • Preview for Next Session: Hands-on hypo testing problems will be undertaken in the next class on Monday, applying the theoretical understanding covered.