(121) JASP 0.10 Tutorial: Paired Samples T-test (Episode 7)

Introduction to Paired Sample T-Test

  • Focus of the video: Paired samples t-test in JASP.

  • Other t-tests previously covered in the tutorial series.

Overview of Data

  • Data file used: Mood and Aggression sample data (available in JASP data library).

  • Data includes 15 observations measuring dementia patients' behavior.

  • Variables:

    • Mood: Average number of disruptive behaviors on full moon days.

    • Other: Average number of disruptive behaviors on other days.

  • Critical for paired-samples t-test: Two measurements from the same subjects.

Null Hypothesis

  • Hypothesis: No difference in average disruptive behaviors between full moon days and other days.

Performing the T-Test in JASP

  • Steps to perform:

    1. Open the data file in JASP.

    2. Select mood and other variables to create pairs.

    3. View results output, including t-value, degrees of freedom, and p-value.

Checking Normality Assumptions

  • Normality is checked for the difference score (moon - other).

  • Shapiro-Wilk test (W value) indicates normal distribution based on p-value.

  • If not normally distributed, Wilcoxon signed-rank test is used instead.

Wilcoxon Signed-Rank Test

  • Nonparametric alternative when the difference score is not normally distributed.

  • Output includes Wilcoxon test statistic and corresponding p-value.

Additional Statistics in Output

  • Statistics that can be included:

    • Mean difference score (moon - other).

    • Standard error for the mean difference.

    • 95% confidence interval for the mean difference.

    • Cohen's d effect size (indicative of the effect strength).

    • Descriptive statistics table (n, means, standard deviations).

    • Descriptive plot with confidence intervals.

Handling Missing Values

  • Strategies for excluding cases with missing data:

    • Analysis by analysis exclusion: excludes missing data only for specific pairs.

    • List-wise exclusion: excludes any subject with missing data from all analyses.

Alternative Hypothesis Selection

  • Three possible alternative hypotheses:

    1. No difference (the null hypothesis).

    2. Two-sided hypothesis (a difference exists without direction).

    3. One-sided hypotheses (specifying greater or less).

  • Choosing two-sided hypothesis results in 2.5% at both tails in t-distribution.

Conclusion

  • Determine whether the effect is significant based on the output results and assumptions.

  • Interpretation depends on the selected hypothesis.