Statistical Inference: Applications, Assumptions, and Limitations

Statistical inference involves using data from a sample to make valid conclusions or generalizations about an entire population. In VCE Specialist Mathematics, this focus is primarily on the population mean (\(\mu\)) using the sample mean (\(\bar{x}\)).

1. The Foundation: Sampling Distribution of the Mean

To apply statistical inference to real data, we must understand the behavior of the sample mean \(\bar{X}\).

• Expected Value: \(E(\bar{X}) = \mu\)
• Standard Error: \(SD(\bar{X}) = \frac{\sigma}{\sqrt{n}}\), where \(\sigma\) is the population standard deviation and \(n\) is the sample size.
• The Central Limit Theorem (CLT): For a sufficiently large sample size (\(n \ge 30\)), the sampling distribution of the mean \(\bar{X}\) is approximately normally distributed, regardless of the population’s original distribution:
  \$\(\bar{X} \approx N\left(\mu, \frac{\sigma^2}{n}\right)\)\$

KEY TAKEAWAY: The Central Limit Theorem is the “bridge” that allows us to use normal distribution techniques on real-world data even when the underlying population distribution is unknown or non-normal.

2. Confidence Intervals for the Population Mean

A confidence interval (CI) provides a range of plausible values for the population mean \(\mu\), based on a sample mean \(\bar{x}\).

Calculation

For a \(C\%\) confidence interval:
\$\(\left( \bar{x} - z\frac{\sigma}{\sqrt{n}}, \bar{x} + z\frac{\sigma}{\sqrt{n}} \right)\)\$

Common \(z\)-values for confidence levels:
| Confidence Level | \(z\)-score (approx.) |
| :— | :— |
| 90% | 1.645 |
| 95% | 1.96 |
| 99% | 2.576 |

Factors Affecting the Width of the Interval

1. Sample Size (\(n\)): Increasing \(n\) decreases the width (more precision).
2. Confidence Level: Increasing the confidence level (e.g., 95% to 99%) increases the width.
3. Standard Deviation (\(\sigma\)): Higher variability in the population increases the width.

EXAM TIP: If an exam question asks you to interpret a 95% confidence interval, avoid saying “There is a 95% probability that \(\mu\) is in this interval.” Instead, say: “In the long run, 95% of intervals constructed using this method will contain the true population mean \(\mu\).”

3. Hypothesis Testing with Real Data

Hypothesis testing is a formal procedure for investigating whether a claim about a population mean is supported by sample evidence.

The Components

1. Null Hypothesis (\(H_0\)): The status quo or the claim to be tested (e.g., \(H_0: \mu = \mu_0\)).
2. Alternative Hypothesis (\(H_1\)): What we suspect might be true instead (\(H_1: \mu \neq \mu_0\), \(\mu > \mu_0\), or \(\mu < \mu_0\)).
3. Test Statistic: The \(z\)-score calculated from the sample:
   \$\(z = \frac{\bar{x} - \mu_0}{\sigma/\sqrt{n}}\)\$
4. \(p\)-value: The probability of obtaining a sample mean at least as extreme as the one observed, assuming \(H_0\) is true.

Decision Rule

• If \(p < \alpha\) (significance level, usually 0.05): Reject \(H_0\). There is statistically significant evidence for \(H_1\).
• If \(p \ge \alpha\): Do not reject \(H_0\). There is insufficient evidence to support \(H_1\).

VCAA FOCUS: Always state your conclusion in the context of the original problem. For example: “Since \(p = 0.03 < 0.05\), we reject \(H_0\). There is evidence to suggest the mean height of the plants has increased.”

4. Errors in Statistical Inference

When applying inference to real data, there is always a risk of reaching the wrong conclusion.

COMMON MISTAKE: Students often think \(P(\text{Type I Error}) + P(\text{Type II Error}) = 1\). This is incorrect. However, there is a trade-off: decreasing the probability of a Type I error (by lowering \(\alpha\)) usually increases the probability of a Type II error.

5. Assumptions and Limitations

When working with real-world datasets, the validity of our inference depends on several key assumptions. If these are violated, the results may be misleading.

Key Assumptions

1. Random Sampling: The data must be a simple random sample. Every member of the population must have an equal chance of being selected.
2. Independence: Observations must be independent of one another.
3. Normality:
   • If \(n < 30\), the population from which the sample is drawn must be normally distributed.
   • If \(n \ge 30\), the Central Limit Theorem allows us to assume the sampling distribution of the mean is normal.
4. Known Standard Deviation: The population standard deviation (\(\sigma\)) is assumed to be known. In practice, if \(n\) is large, the sample standard deviation (\(s\)) is often used as an estimate.

Limitations in Real Data

• Sampling Bias: Real data is often “convenience data.” If the sample is biased (e.g., only surveying people at a library about literacy), the inference cannot be generalized to the whole population.
• Outliers: Real data often contains outliers that can significantly skew the sample mean \(\bar{x}\), leading to incorrect \(z\)-scores and \(p\)-values.
• Non-Response Bias: In surveys, people who choose not to respond may have different characteristics than those who do, affecting the mean.
• Causality vs. Correlation: Statistical inference can show that a mean has changed or that a relationship exists, but it cannot prove why it happened (causation) without a controlled experiment.

STUDY HINT: When asked to “critique” or “discuss limitations” of a statistical study in an exam, look for mentions of “small sample sizes” (\(n < 30\)), “voluntary surveys” (bias), or “non-random selection.”

6. Summary Table: Application Steps

REMEMBER: “If the \(p\) is low, the null must go!” (If \(p < \alpha\), reject \(H_0\)).