COMPASS

COrrelation Methylation Platform for Association Search and Synthesis

Pooled correlation (Spearman's ρ)
This is the overall strength of association estimated by combining the correlation results from multiple datasets. A positive value indicates a correlation in the same direction, while a negative value indicates an inverse relationship. Because this tool pools Spearman's rho (ρ) from each dataset, the final pooled estimate can also be interpreted as Spearman's rho.

95% CI (95% confidence interval)
This indicates the uncertainty around the pooled correlation estimate. A wider interval implies greater uncertainty. If the interval includes 0, it is difficult to conclude that there is an overall correlation.

P-value (overall effect)
This tests the null hypothesis that the overall correlation is zero (i.e., no association). Smaller values indicate stronger evidence for an overall association.

Studies (k)
The number of datasets (studies) that were actually included in the meta-analysis (datasets that could not be used due to missing data, etc., are not counted).

Heterogeneity (between-dataset variability)

I²
A measure (percentage) of how much the results vary across datasets. Higher values suggest that the estimated correlations are less consistent across datasets, potentially due to differences in cohorts, measurement conditions, preprocessing, and other factors.

τ² (tau-squared; between-study variance)
This quantifies the magnitude of between-dataset variability (as a variance). It is less intuitive than I², but it represents the amount of dispersion in the "true" effects assumed by the random-effects model. Because the meta-analysis is performed on a transformed scale (for pooling correlations), τ² is the between-study variance on that transformed scale.

P-value (heterogeneity / Q test)
This tests the null hypothesis that there is no heterogeneity (i.e., all datasets share the same true effect). Smaller values indicate that between-dataset differences may be larger than expected by chance.

Weight
The relative contribution of each dataset to the pooled result. In general, datasets with larger sample sizes and more precise estimates receive greater weight.