Module 06 Summary

What this module established

Pairing subtracts the person out. With a between-person SD of 12.77 and a within-person change SD of 2.83, the unpaired test found nothing (p = 0.66) and the paired test found a clear effect (p = 0.008) with an interval 6.4 times narrower.

Carry forward

• Welch is R's default and is the safer choice. With unequal spreads and unequal group sizes it gave 14.6 degrees of freedom where the equal-variance test claimed 73.
• A chi-square test detects association and explains nothing. The standardised residuals showed premium churning at +3.04 and basic at -2.72.
• After a significant ANOVA, adjust the pairwise comparisons. Only one of three pairs survived Tukey's HSD.
• Normality tests are least useful where you need them: 38% detection of genuine skew at n = 8. And a real 5-point effect is missed 90% of the time at n = 10 per group.

Before moving on

Move on when you can state, for any null result you report, what effect the study could have detected.

Mark complete