Getting NaN as the mean of a row that contains no NaN values, Inf values, 0/0 values

Dear MATLAB experts,

I'm trying to calculate the mean values of each row of a table, but I keep on getting NaN as the value of the mean of each row.

I don't understand why I'm getting this error, since all values are doubles and there are no zero values / infinite values in the dataset. My code is the one below. I've also attached capmEstimates90 to this post.

capmEstimates90 = table2array(capmEstimates90)
capmEstimates90Stats = [];

% Calculating statistics
for i=1:height(capmEstimates90)
    
    capmEstimates90Stats(1,i) = mean(capmEstimates90(:,i));
    
end

I would really appreciate your help. Thank you in advance.

[Alberto Cuadra Lara]

I do not obtain NaN values, but you can exclude Nan values using omitnan.

capmEstimates90 = table2array(capmEstimates90);
% Calculating statistics
for i=length(capmEstimates90(:,1)):-1:1 % Preallocate variable
    capmEstimates90Stats(1,i) = mean(capmEstimates90(:,i), 'omitnan');
end

Answer 1

[DGM]

Not sure if that's really supposed to be a wide table. You say you're calculating row means, but you're calculating column means instead. If your table is supposed to be transposed, then just transpose it. At any rate, there's no need for loops.

load capmEstimates90.mat

capmEstimates90 = table2array(capmEstimates90);

% let's say you want several different stats (examples)
columnmin = min(capmEstimates90,[],1); % min of each column
columnmax = max(capmEstimates90,[],1); % max of each column
columnmeans = mean(capmEstimates90,1); % mean of each column
columnnans = sum(isnan(capmEstimates90),1); % number of NaNs in each column
columnmeans2 = mean(capmEstimates90,1,'omitnan'); % mean of each column, ignoring NaN

% if you want to put them in one big array
outputstats = [columnmin; columnmax; columnmeans; columnnans; columnmeans2];

Yes, there are NaNs in there.

numberofnans = nnz(isnan(capmEstimates90))

numberofnans = 92

If you're truly doing things by row instead of by column, contrary to the code you posted, then yes, each row has at least one NaN.

nansperrow = sum(isnan(capmEstimates90),2)

nansperrow = 6×1

8 8 19 19 19 19

Though if you're doing things by column as described, most columns contain no NaNs, and no column contains more than six NaNs.

[counts edges] = histcounts(sum(isnan(capmEstimates90),1),'binmethod','integer');
nanspercol_instances = [edges(2:end)-0.5; counts].'

nanspercol_instances = 7×2

0 1851 1 0 2 0 3 0 4 11 5 0 6 8