Hi Joe,

Could you confirm the default hierarchy of layering of subjects, conditions and electrodes in the arrays given in the sub-panes of the PCA window? This may be in the documentation but I haven’t spotted it. I’m planning on using the data in some of these sub-panes in a macro to be run on ERPs in Brain Vision Analyzer. To do this I need to know what combination of subject, condition and electrode each row in the sub-panes refers to

For example in FacScr, with 10 subjects, 64 electrodes and 4 conditions, will the first 64 rows the scores of Subject 1 Condition 1, Electrodes 1-64, the next 64 rows Subject 1 Condition 2 Electrodes 1-64? That is: is the hierarchy Subject, Condition, Electrode?

Although some preliminary investigations suggested that was the likely layering, I found some inconsistencies when I tried to reproduce temporospatial factor scores from the data in other sub-panes. In the following example there is just a single subject in the PCA, 4 conditions and 64 electrodes . To re-create the factor scores for TF1SF1 that I found in FacScrST I took the 256 factor scores for TF1 from FacScr and standardised them. I then applied the 64 coefficients of TF1SF1 found in FacCoffST to the first 64 TF1 factor scores and summed the products to generate a TF1SF1 factor score for condition 1 (assuming I got the layering hierarchy right). I applied the same 64 coefficients to TF1 factor scores 65 to 128 to get the TF1SF1 factor score for condition 2 and so on. However the four resulting TF1SF1 scores are very slightly adrift from those reported in FacScrST, consistently showing larger absolute values.

It’s possible that this just shows I have misconceived the spatial step of the PCA. Referring to your chapter in Handy you say the spatial step is achieved by “rearranging the factor scores resulting from the temporal PCA so that each column contains the factor scores from a different electrode. A spatial PCA can then be conducted using these factor scores as the new variables” However, it’s not clear to me how the data matrix for the spatial step can have factor scores as variables in the strict sense of their being column headers. Instead I have conceived of the data matrix (in the present example) as 64 columns corresponding to the electrodes, and four rows corresponding to the observations (since there is only one subject) with the contents of the data matrix being TF1 scores for electrode x condition.

Thanks in advance

Tom