A contrast is an instantiation of a hypothesis. The omnibus contrast is a test for any task-related activity against the baseline. Here, the “baseline” means the variance left unexplained after fitting the time course at each voxel for the mean, linear gtrend, and task-related activity. Mathematically, the mean of the baseline is forced to zero. The omnibus is usually tested first to ensure that the data have been processed properly and that the subject was responding. If there is no omnibus activity, then there is no use in continuing the analysis.

As with mkanalysis-sess, this step merely prepares the required information for the computation, which actually occurs later. A contrast only needs to be made ONCE. If you change the analysis (i.e. the time window, the prestimulus window, or the number of conditions), then you have to recreate the contrast.

You will supply a name for this comparison, which will be required in subsequent processing steps. For multiple comparisons, you would run mkcontrast-sess multiple times, defining a different contrast name each time. The contrast is defined with respect to a particular analysis by using the command options.

In your Study Directory, you would run mkcontrast-sess with the all the following options:

-analysis analysisname

name of analysis

-contrast contrastname

name of contrast (you have to name it, e.g. “omnibus”)

-a active_condition_code (#)

to specify >1 active conditions use multiple –a options

-c control_condition_code (#)

to specify >1 active conditions use multiple –c options


specifies that the conditions be tested separately

For Bert, type:

mkcontrast-sess -analysis sem_assoc -contrast omnibus -a 1 -a 2 -a 3 -a 4 -c 0 -nosumconds

OUTPUT: This will create a file called “omnibus.mat” under the “analysisname” directory in your Study Directory:

Condition codes are the same as those used in the paradigm file specified for this analysis in mkanalysis-sess.new.

FsFastTutorialV4.5/910_Define_an_omnibus_contrast (last edited 2011-05-19 15:29:08 by NickSchmansky)