|
Abstract:
|
Data for behavioral and neuroimaging studies tend to be challenging due to the nature of heterogeneity, autocorrelation and higher level of noise; moreover, the multidimensional feature of neuroimaging is aggravated by structural and functional correlations; and all these raise the challenge to a new high. Under these circumstances, unsuitable statistical methods and tools combined with marginalized effects are more likely to take place; and inflated errors and low power are the common collective results of the analyses. As a consequence, scientific rigor, power analyses and reproducibility have become a very serious concern in mental health studies. In this presentation, we would like to identify and discuss the gaps between the current statistical methods and the desired demand for appropriate data analyses by emphasizing the importance of alternative statistical methods and tools to narrow down the gaps and improve the scientific rigor and reproducibility.
|
