The codes take two variables x and y, where x is a n by p matrix consisting n subjects and p features, and y is a n by 1 vector consisting behavior measurement. The codes then conduct (1) feature selection, and (2) leave-one-subject-out cross-validation using selected features of x on predication of y, as detailed below.
|License:||GPL (>= 2)|
Pos(itive) refers to the positive network. A node was considered a part of the positive network if it was positively (and significantly) correlated with the behavior measurement. More specifically, when conducting feature selection during the training stage using data from n-1 subjects, we found the correlation of each node and the behavior (hence p, where p = number of features/nodes, correlations), and considered features/nodes with significant positive correlation as a part of the positive network. The selected (positive) features/nodes were then used to predicate behavior data in a holdout training set. This was conducted iteratively for n times. Finally, we took the correlation between the predicated and observed behavior to assess the power of the model. The neg(ative) approach was conducted similarly.
Oliver Y. Chén
Maintainer: Oliver Chén <firstname.lastname@example.org>
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.