Introduction to T-Tests: A Statistical Cheat Sheet
The t-test is a statistical inference tool used when the population standard deviation is unknown, allowing researchers to determine if sample means significantly differ from a hypothesized population mean by accounting for the increased uncertainty of estimating variability from sample data.
Core Principles
- NHST Logic: Reject the null hypothesis (H0) if the observed sample statistic is sufficiently unlikely under H0, meaning it falls beyond a critical value determined by alpha (α).
- Z-Test Limitation: Assumes knowledge of the population standard deviation (σ), which is often unavailable in psychological research.
- T-Distribution: A bell-shaped, symmetrical distribution with heavier tails than the normal distribution, accounting for the added uncertainty when estimating σ from sample data (s).
- Degrees of Freedom (df): Calculated as n-1, df adjusts the t-distribution's shape based on sample size; larger df makes the t-distribution approach the normal distribution.
- T-Test Formula: T_observed = (M - μ₀) / s_M, where s_M is the estimated standard error of the mean (s/√n).
Action Steps
- 1. State Hypotheses: Define the null (H0: no difference) and alternative (HA: a difference exists) hypotheses.
- 2. Determine Rejection Region: Identify the critical t-value (T_critical) using alpha (α), degrees of freedom (df), and whether the test is one-tailed or two-tailed.
- 3. Compute Test Statistic: Calculate the observed t-value (T_observed) using the sample mean (M), hypothesized population mean (μ₀), and the estimated standard error of the mean (s_M).
- 4. Make a Decision: Compare T_observed to T_critical. If |T_observed| > |T_critical|, reject H0; otherwise, fail to reject H0.
Formulas
- Z-observed = (M - μ₀) / (σ/√n)
- T_observed = (M - μ₀) / (s/√n)
- s_M = s/√n
- df = n - 1
Key Terms
- Standard Error of the Mean (SE_M): Measures the expected variability of sample means around the population mean; calculated as σ/√n for z-tests and s/√n for t-tests.
- T-distribution: A probability distribution used in hypothesis testing when the sample size is small or the population standard deviation is unknown, characterized by heavier tails than the normal distribution.
- Degrees of Freedom (df): The number of independent values that can vary in the data; for a single sample t-test, it's the sample size minus one (n-1).
- Estimated Standard Error (s_M): The standard error of the mean calculated using the sample standard deviation (s) instead of the population standard deviation (σ).
Pro Tips
- Use a one-tailed t-test when your alternative hypothesis specifies a direction (e.g., 'higher than' or 'lower than').
- Use a two-tailed t-test when your alternative hypothesis states a difference exists but does not specify direction (e.g., 'different from').
- As sample size (n) increases, the t-distribution more closely resembles the normal (z) distribution because the estimate of population variability (s) becomes more reliable.
Pitfalls to Avoid
- Using a z-test when the population standard deviation (σ) is unknown leads to an inaccurate assessment of statistical significance due to unaddressed uncertainty.
- Misinterpreting 'fail to reject H0' as 'accept H0' incorrectly concludes no difference exists when evidence is simply insufficient.
Myth vs Reality
- The t-distribution is identical to the normal distribution.: The t-distribution has heavier tails than the normal distribution to account for the additional uncertainty introduced when estimating the population standard deviation from sample data.
Real World Examples
- Testing if adolescents whose parents allow sipping have higher drinking intentions than the neutral midpoint (4) on a 1-7 scale.: A one-tailed t-test was used with M=4.60, s=1.20, n=25. The calculated T_observed (2.50) exceeded the T_critical (1.711), leading to rejection of the null hypothesis (μ=4) and concluding significantly higher intentions.
- Testing if students at UB score higher than the national average (μ=100) on an intelligence test, knowing population σ=15.: A one-tailed z-test was performed with n=36. The calculated Z_observed (2.00) exceeded the Z_critical (1.645), leading to rejection of the null hypothesis (μ=100) and concluding UB students score significantly higher.
Statistics
- Alpha level (α) for critical value determination: Typically set at 0.05
- Degrees of freedom (df) calculation: df = n - 1
- T-critical for a one-tailed test (α=0.05, df=24): 1.711
- T-critical for a two-tailed test (α=0.05, df=24): ±2.064
People
- William Sealy Gosset: Chemist for Guinness Brewery, developed the t-distribution and t-test to address small sample size issues in experimental data.
Related cheat sheets
More like this