Statistical testing of images and time courses


Two Sample t-test of Component Images

Two sample t-test can be used to compare a component image between subjects or sessions. For example for a particular component image session 1 can be treated as group1 and session 2 as group2. You can use these images in SPM to do two sample t-test ([3], pp 12-14).

Regression Parameters

After performing temporal sorting, the regression parameters are saved in a file with suffix "regression.txt". This file can be quite useful in performing statistical tests to evaluate the task-relatedness of the components or to perform tests between components or groups. The text  file contains a number for each component, regressor, subject and session (Table 1). The numbers represent the fit parameter from the regression. In order to use these numbers it is important to scale the ICA data. For example, if 10 subjects were analyzed with ICA and two regressors were used to sort, then one can test the degree to which the average amplitude of group 1 was greater than the average amplitude of group 2 by computing a two-sample t-test between the 10 parameters for group 1 and the 10 parameters for group 2. A separate test can be done for each component and the component which show a significant difference can be determined.

Table 1: Table shows two sorted components parameters. Regression values for each component are shown followed by the beta weight of each condition with the ICA time course.

Component Numbers 14 13
Regression Values 0.82740899 0.74477757
Subject 1 Sn(1) left*bf(1) -0.052490765 1.6484661
Subject 1 Sn(1) right*bf(1) 1.4887159 0.13184605
Subject 2 Sn(1) left*bf(1) 0.12544365 1.0076309
Subject 2 Sn(1) right*bf(1) 1.6412895 0.75554373
Subject 3 Sn(1) left*bf(1) -0.35264506 1.5612781
Subject 3 Sn(1) right*bf(1) 1.3959309 -0.0077817872

 We have now provided GUI for doing statistics on time courses (beta weights) under Utilities drop down box.