details concerning module "AutoANOVA"

07/10/2010 09:16

Some information concerning the auto-anova. 

1. I have double-checked. The output generated by the Letswave 4 autoanova is identical to the output generated by the Letswave 3 autoanova, which is reassuring. Note that given the large amount of calculations to generate this anova, I had verified that the F values generated by the autoanova were identical to those generated by SPSS. 

2. How to use the Letswave 4 autoanova module: 
- Input files should contain n epochs, each corresponding to a different subject, and each input file should correspond to a different condition. 

- In the upper panel, you define the structure of your ANOVA, which can include up to 2 between-subject factors and 2 within-subject factors, with as many levels as you want per factor. In Diana's experiment, we have two within-subject factors (certainty (2 levels) and difference (2 levels)). 0 defines the first level, 1 the second, etc. 

- In the lower panel, you define which results you want to include in the output file. In our case, we would like to have the main effect of certainty, the main effect of difference and the interaction between the two factors. So, you will need three epochs to output your results. For the first two, you will select a single factor (main effect of that factor), for the last one, you will select the two factors (interaction between those two factors). 

Hence, the window should look like this : 

- After you run the auto-anova, you will be given two files : the auto-anova_Fvalues, which contains the F values, the auto-anova Pvalues which contains a series of 0s and 1s (0=not significant, i.e. p >=0.05; 1=significant, i.e. p <0.05). In our example, each file will contain 3 epochs. The first epoch will correspond to the main effect of factor within1, the second epoch will correspond to the main effect of factor within2, the third epoch will correspond to the interaction within1xwithin2. 

3. Consecutivity threshold.

You may want to apply an arbitrary consecutivity threshold to your results. I would perform this using the Pvalues files (the file with the 0s and 1s). Set your threshold to "greater or equal" than "1". Then set your consecutivity threshold to the desired value (e.g. +/-5 bins). This will generate a new Pvalues 
file, where only the time points that were surrounded by a sufficient number of "significant bins" will be retained. Finally, for display purposes, you may want to multiply (using the Arithmetic Operations using pairs of datasets module) your Fvalues file with your Pvalues file, to retain only those Fvalues that were above the threshold.