Table 1.

Summary of the quantity of information found by type of data.

	Methods and Tools	Most Frequent Strategy Used
Within-Subject Correlation	To quantify intraclass correlation: Modifications of Pearson’s product-moment correlation coefficient. Comparisons based on the generalized estimating equations generated by mixed-effect models.	Analysis of data using a mixed-effect linear model that can accommodate a dependent variance-covariance structure.
Multiplicity	Controlling the family-wise error rate (Tukey, Bonferroni, Scheffe and other). Approach to controlling the false discovery rate used in biomarker studies: Benjamini and Hochber.	Analysis of data using a methodology that controls the family-wise error rate.
Multiple Clinical Endpoints	The selection of a single primary endpoint for formal statistical inference, considering that the endpoints are possibly biologically related and positively correlated. Creating a univariate outcome by combining multiple clinical endpoints (weighted measures taking into account the relevance of each endpoint). To compare the two samples based on the endpoint of highest priority first, and, if no winner can be determined, would one move to the endpoint of the next highest priority.	Analysis of data by prioritizing the relevant endpoints or by using a composite endpoint.
Selection bias	To adjust for age, stage, treatment and so forth. Matched samples.	Analysis of data using a multivariate model to simultaneously adjust for confounders Obtention of matched samples. Propensity score weighted.
Publication bias	To publish positive and negative results.	Encourage the objective assessment of molecular signatures by reporting both positive and negative outcomes. Make data publicly available after publication.