p値じゃなくeffect estimate values (e.g., percent signal change, correlation)がいんじゃね

Is the Statistic Value All We Should Care about in Neuroimaging?
Neuroimage. 2016 Oct 8;:
Authors: Chen G, Taylor PA, Cox RW

1)
the statistic values be used for thresholding only (not for colorization, determining maxima of activity, etc.);
2)
the activation patterns in brain images be colored by effect estimate values (e.g., percent signal change, correlation), not by statistic values; and
3)
the full set of parameters (threshold value, degrees of freedom for each statistic test, cluster-wise probability, etc.) be explicitly stated.

を推奨。