Alexis Pérez's reported bugs

[caiw]

The following was submitted to me some time ago by Alexis Pérez of Baylor College of Medicine, and I never had a chance to go over it in detail.

Alexis suggested some fixes, but we should verify that they function as intended.

This was reported by Alexis in response to me sending him a file to fix another bug ( issue #18 ).

I was trying to run the analyses adding the new function that you sent me. I found some other problems and I am not sure if the errors are produced by the way my data are set or these are the consequences of some RSA toolbox bugs. Here below I pass you a list with some of these problems. Maybe you can use this to improve the RSA toolbox functionality.

Using the Randomisation test,

In compareRefRDM2candRDMs, I have to set the number of randomisations to a specific number. For example, my data contain 5 different stimuli classes, so if I do not set the n of randomisations below 121 the program reports a problem with the matrix dimensions.

Attempted to access
allPermutations(121,:); index out
of bounds because
size(allPermutations)=[120,5].

After correcting the number of randomisations, the new problem appearing is that the function is not able to find the es variable.

Undefined function or variable
"es".

Error in compareRefRDM2candRDMs
(line 727)
stats_p_r.SEs=es;

Plotting horizontal bar for Canddifferencestest

In compareRefRDM2candRDMs, if there are not significant differences in the Candidatedifferencestest , the threshold variable will be empty, thus, because the program tries to draw the horizontal bar and there is no yy variable (a given threshold is necessary).

Some related problems are present when the program goes through the addComparisonBars function. Specifically, in line 31

if thisPairwiseP <= threshold

if thisPairwiseP variable is larger than the threshold variable, then matlab reports the next error

Error in addComparisonBars (line 13)
nTestRDMs = size(pairwisePs,1);

Output argument "y" (and maybe others) not
assigned during call to
"addComparisonBars".

Error in compareRefRDM2candRDMs (line 886)
yy=addComparisonBars(stats_p_r.candDifferences_p,(ceilingUpperBound+0.1),threshold);

FDR correction

In compareRefRDM2candRDMs, if I set candRDMdifferencesMultipleTesting = 'FDR'; when correcting the Candidatedifferencetest pvalues for FDR, if there are no significant effects, FDRtreshold returns an empty variable. Condequently, when running
u(pMat>thr_fdr) = 1;% black

matlab reports an error.

I have solved this problem by:

if isempty(thr_fdr)

thr_fdr = 1

end

Bootstrap function
I have also tried to use the bootstrap function. It usually takes around 4 hours with 10000 iterations. Now I dont have the code output at hand but matlab also reported a problem with the plotting functions. The next week I will perform some test with the bootstrap functions, possibly I can give you some feedback.

I can more or less modify the code by myself (I am a little bit messy but...) in order to finish my analyses, anyways I hope my comments could be useful to correct possible bugs.

I can pass you my datasets and my code if you want to do some tests too.
My data consist of

10 subjects
5 conditions
8 ROIs
2 Models