Considered together, the above mappings allowed us to assign a set of voxels (from different lobes and participants) to each of the second-level factors.
(The percentage of variation accounted for by each of the four second-level factors in the data of the four participants with plentiful anterior voxels was eating: 7.26; shelter: 8.51; manipulation: 7.31; word length: 5.67).
Similarly, for the second level analysis, a first-level factor was uniquely assigned to one of the 10 second-level factors for which it had the highest (absolute value) loading, provided that this loading has was above the 0.4 threshold.
The input to the second-level analysis consisted of the five dominant first-level factors obtained from each lobe of each participant.
The five dominant factors were selected from each of these first-level analyses, to produce a set of 100 first
The first four factors (explaining 20% of the variation in the first-level factors data) were extremely similar to the corresponding factors from the original four-participant analysis, and also were shared by a substantial proportion of the participants.
For the first-level analyses, a voxel was uniquely assigned to one of the five first-level factors for which it had the highest (absolute value) loading, provided that this loading was above a threshold value of 0.4 (a typical value for exploring factor structure).
Solving the above equation for the unknown second-level factor profiles (using least squares) produces the vectors of second-level factor profiles over the 60 words.
Of the factors emerging in the second-level factor analysis, only four of them were common to all four of the participants with plentiful anterior voxels in the first-level factor analyses.
Then a second-level factor analysis was run to identify factors that were common across lobes and participants, a procedure known as higher-order factor analysis [16].
This algorithm was applied separately for each set of 50 voxels selected from five lobes of four participants, resulting in 20 first-level factor analyses.
