The ANCOVA found that the group effect is statistically significant after controlling for the covariate (F = 91.657, p < 0.001). The partial eta-squared for the group factor is 0.0842, indicating a medium effect size. The model explains 67.4% of outcome variance (R² = 0.6742). The highest covariate-adjusted group mean belongs to 'completed' (71.63).

The table shows raw group means and standard deviations for both the outcome and the covariate. Outcome means range from 64.5 to 74.42 across the 2 groups. Covariate means range from 64.08 to 69.7, indicating notable baseline differences. Where groups differ substantially on the covariate, ANCOVA adjustment will meaningfully shift the group means.

The ANCOVA table uses Type II sums of squares, testing each predictor after adjusting for the other. The group factor is statistically significant (F = 91.657, p < 0.001) with partial η² = 0.0842. Partial eta-squared thresholds: ≥0.01 small, ≥0.06 medium, ≥0.14 large. A covariate partial η² above the group factor indicates that controlling for it substantially reduced error variance.

Partial eta-squared (η²) quantifies the proportion of outcome variance explained by each predictor after removing the other's contribution. The largest effect belongs to 'Covariate' (η² = 0.6388). Comparing covariate vs group effect sizes reveals whether prior differences or the treatment itself drives more of the outcome variation.

Adjusted (least-squares) means represent what each group's average outcome would be if all groups had identical covariate values (grand mean). Error bars show 95% confidence intervals. The groups' confidence intervals do not overlap, providing strong evidence of a real difference. The gap between 'completed' (71.63) and 'none' (66.06) is the covariate-adjusted effect.

Each point is one observation, coloured by group. The scatter shows the covariate-outcome relationship (r = 0.803 across all observations). Approximately parallel regression lines across groups support the homogeneity-of-slopes assumption required for valid ANCOVA. If lines cross or fan out, an interaction term (group × covariate) would be needed.

Each group has two bars: the raw (unadjusted) mean and the ANCOVA-adjusted mean. The adjusted mean corrects for the grand-mean covariate level, isolating the pure group effect. The largest shift belongs to 'completed' (Δ = 2.79 points), indicating the covariate was confounding that group's raw average the most. Groups with small shifts were already near the grand-mean covariate level.

The histogram displays the distribution of ANCOVA model residuals (mean = 0, SD = 8.67). A bell-shaped distribution centred near zero confirms the normality assumption. Shapiro-Wilk p = 0 — normality rejected; interpret with caution. Heavy tails or skew suggest that F-test p-values and confidence intervals may be slightly liberal.