Table 1.

Summary of advantages and disadvantages of various mass univariate methods.

Method	Advantages	Disadvantages
Permutation-based Fmax correction	Best power for spatially and temporally focal ERP effects. Controls the probability that even one false positive time point is present, allowing for claims that each individual significant time point represents a true effect.	Less power for spatially and temporally extended effects, especially if the effect is not large at its peak. Substantially underestimates the true temporal extent of effects.
Permutation-based cluster mass correction	Best power for spatially and/or temporally broadly distributed effects. When overall power is high, gives a reasonable estimate of the time course of effects.	Does not allow for claims about whether individual time points show an effect with a given error rate. When overall power is low, clusters may include many false positive time points.
False Discovery Rate correction (Benjamini & Hochberg, 1995; Benjamini, Krieger, & Yekutieli, 2006)	Can be combined with any statistical model or test conducted at each time point/electrode and thus extendable to models that are not feasible with permutation tests (e.g. single trial mixed linear regression). Provides reasonable power to detect effects, albeit less than the permutation-based methods.	Less power than permutation-based cluster methods to detect extended effects, and less power than permutation-based Fmax methods to detect focal effects. Statistical assumptions may not be met by EEG data, leading to an inflated false discovery rate at individual time points.
False Discovery Rate correction (Benjamini & Yekutieli, 2001)	Can be combined with any statistical model or test conducted at each time point/electrode and thus extendable to models that are not feasible with permutation tests (e.g. single trial mixed linear regression). Makes no assumptions about correlation between time points and electrodes, and thus correctly controls false discovery rate at individual time points.	Generally offers the lowest power of all methods.