Independent Samples T-Test Cheat Sheet
The independent samples t-test compares the means of two independent groups to determine if the observed difference is statistically significant or due to random chance, using pooled variance when assumptions are met and effect sizes to quantify the magnitude of the difference.
Core Principles
- Independent samples t-test assumptions: Interval/ratio data, independent observations within samples, random sampling, normality (large samples), and homogeneity of variance (equal population variances).
- Pooled variance is calculated when sample variances are similar or sample sizes are unequal to combine variability estimates.
- Effect size (Cohen's d or r-squared) quantifies the magnitude of the difference between groups, independent of statistical significance.
- Confidence intervals provide a range of plausible values for the true population difference between means, accounting for sampling variability.
Action Steps
- 1. State null (H0) and alternative (H1) hypotheses regarding population means.
- 2. Check assumptions: interval/ratio data, independence, random sampling, normality, and homogeneity of variance.
- 3. Test for homogeneity of variance: If larger s²/smaller s² ≥ 2.00, variances are unequal; otherwise, they are considered equal.
- 4. Calculate pooled variance (sₚ²) if variances are equal or sample sizes are unequal.
- 5. Calculate the pooled standard error (s<0xE1><0xB5><0x83>₁₋<0xE1><0xB5><0x83>₂).
- 6. Calculate the observed t-statistic (t observed).
- 7. Determine the critical t-value (t critical) based on alpha level and degrees of freedom (df = n₁ + n₂ - 2 for pooled variance).
- 8. Make a decision: Reject H0 if |t observed| > t critical; otherwise, fail to reject H0.
- 9. Calculate effect size (Cohen's d or r²) to interpret the magnitude of the difference.
- 10. Construct a confidence interval for the difference between population means.
Formulas
- Pooled Variance: $$s_p^2 = \frac{(s_1^2 \times df_1) + (s_2^2 \times df_2)}{df_1 + df_2}$$
- Pooled Standard Error: $$s_{M1-M2} = \sqrt{s_p^2 (\frac{1}{n_1} + \frac{1}{n_2})}$$
- Observed T-statistic: $$t = \frac{(M_1 - M_2) - (\mu_1 - \mu_2)}{s_{M1-M2}}$$
- Cohen's d: $$d = \frac{|M_1 - M_2|}{s_p}$$
- Variance Explained (r-squared): $$r^2 = \frac{t^2}{t^2 + df}$$
- Confidence Interval: $$CI = (M_1 - M_2) \pm (t_{critical} \times s_{M1-M2})$$
Key Terms
- Independent Samples T-Test: Compares the means of two independent groups to determine if their population means differ significantly.
- Homogeneity of Variance: The assumption that the population variances of the two groups being compared are approximately equal.
- Pooled Variance: A weighted average of the variances of two independent samples, used when assuming equal population variances.
- Pooled Standard Error: An estimate of the standard deviation of the sampling distribution of the difference between two independent sample means, calculated using pooled variance.
- Effect Size: A measure that quantifies the magnitude of the difference between two groups, independent of sample size.
- Cohen's d: An effect size measure representing the difference between two means in standard deviation units.
- Variance Explained (r²): An effect size measure indicating the proportion of total variance in the dependent variable that is associated with group membership.
- Confidence Interval (CI): A range of values that is likely to contain the true population parameter (e.g., the difference between two population means) with a certain level of confidence.
Pro Tips
- Pooling variance is recommended even if variances are equal when sample sizes are unequal to improve the estimate of population variability.
- If variances are unequal (ratio ≥ 2.00), use unpooled standard error and a conservative, smaller df (the minimum of n₁-1 or n₂-1) for a more conservative test.
- When interpreting results, consider both statistical significance (p-value from t-test) and practical significance (effect size).
Pitfalls to Avoid
- Assuming equal variances when the ratio of larger s² to smaller s² is 2.00 or greater leads to inaccurate standard error and potentially incorrect conclusions.
- Ignoring effect size can lead to overemphasizing statistically significant but practically meaningless differences.
- Using a one-tailed critical value when the hypothesis is two-tailed inflates Type I error rate.
Myth vs Reality
- The t-test only tells you IF a difference exists between groups.: While the t-test indicates statistical significance, effect size measures like Cohen's d and r-squared quantify the magnitude of that difference, providing crucial context beyond just significance.
- You must always use pooled variance if sample sizes are different.: Pooling variance is recommended if sample sizes are unequal, but if homogeneity of variance is violated (ratio ≥ 2.00), unpooled standard error should be used instead.
Real World Examples
- Comparing perceived social skills scores between adolescents who participated in the PEERS intervention (Group 1) and those in a control group (Group 2).: An independent samples t-test was used to determine if the PEERS intervention led to significantly higher perceived social skills (M₁=78, s₁²=121) compared to the control group (M₂=70, s₂²=100).
- Evaluating the difference in exam scores between males and females.: An independent samples t-test can assess if there is a statistically significant difference in average exam scores between male and female students, assuming scores are interval/ratio data and participants belong to only one group.
Statistics
- Cohen's d interpretation: Small effect: d < |0.2|
- Cohen's d interpretation: Medium effect: |0.2| < d < |0.8|
- Cohen's d interpretation: Large effect: d > |0.8|
- Variance Explained (r²) interpretation: Small effect: r² = 0.01
- Variance Explained (r²) interpretation: Medium effect: r² = 0.09
- Variance Explained (r²) interpretation: Large effect: r² = 0.25
People
- Nate Caluda-Perdue: Instructor/Presenter for Psychological Statistics 207C
Related cheat sheets
More like this