Issue while estimate number of independent components

[facazevedo]

I was trying to estimate the number of IC components using the "Standard ICA/IVA" option and the MDL (i.i.d. Sampling) but I got an error "Unrecognized function or variable "dim_n". Upon investigating the issue, I found out that the problem lied in the snippet of the function icatb_estimate_dimension.m:

%% Perform variance normalization
if verbose
    fprintf('\n Performing variance normalization ...');
end

for n = 1:size(data, 2)
    data(:, n) = detrend(data(:, n), 0) ./ std(data(:, n));
end

%% (completed)

What happened was that a few of the columns in my dataset had zero std, which made the data to contain columns of nans, which causes svd to fail later. To circumvent that, I added two lines of code that removed the columns of nan's and redefined tdim:

%% Perform variance normalization
if verbose
    fprintf('\n Performing variance normalization ...');
end

for n = 1:size(data, 2)
    data(:, n) = detrend(data(:, n), 0) ./ std(data(:, n));
end

data = data(:, ~all(isnan(data))); % modfred
tdim = size(data, 2); % modfred

%% (completed)

The code seems be able to estimate the PC's but I get the error at the end:

121 eigen values are less than or equal to machine precision

Best.