Calculating Mean of Matrices

I am new to Matlab. I would like to calculate mean of many matrices(about 550 matrices. They are all the same size). I've named them loc1, loc2, loc3 .......loc549, loc550. Please help. Thanks!

[Oleg Komarov]

How did you name them loc1, loc2 etc...manually or did you use eval?

[John]

I used eval to name them. They are all 500x1 matrices and I would like to get the means of their rows.

[Oleg Komarov]

I think you're missing the point, "evaling" variables is not a good approach. You will find yourself stuck in the eval paradigm forever.

Try the other methods suggested in the FAQ that per isakson posted.

[Oleg Komarov]

Sorry but 500x1 are vectors, not matrices and their row-means are equal to the values itselves.

Answer 1

[per isakson]

It was probably not a good idea to name them "loc1, loc2, loc3 .......loc549, loc550". See How can I create variables A1, A2,...,A10 in a loop?.

However,

    mean(loc1(:))

returns the mean of a numerical array.

Please, have a look at the Getting Started.

[John]

Thanks for the advice. If I rename them A1,A2.....how can I calculate the means of their rows (they are all 500x1 matrices )

[Oleg Komarov]

Use the structure approach or do the variables have all the same dimensions?

[John]

They all have the same dimensions

[E K]

try to use a cell structure with each cell element got one of your matrix then you can call them like C{1,1}

and with a for loop it will be very easy to find the mean of them.

[per isakson]

@John

1. you did not read the FAQ-entry, "A1,A2...", carefully enough
2. read the documentation on MEAN

Answer 2

[Oleg Komarov]

Then create one array loc, which will have the usual n by m dimension but you will be adding slices in a loop on the third dimension.

An example, suppose you are importing data (but I will be generating it):

% preallocate
A = zeros(2,4,10); % You know you will have 10 matrices
for ii = 1:10
    A(:,:,ii) = rand(2,4); % one matrix per slice
end

Try to apply this example to your case, then you just have to use

mean(A,2)

mean along rows.

Answer 3

[Azzi Abdelmalek]

 for k=1:550
    mat=sprintf('loc%d',k);result1=sprintf('result(%d,:)',k)
    %eval([mat '=rand(2,2)']) %it's just an exemple for testing
    eval([result1 '=mean(' mat ,')'])
 end

% the variable result is an array with 550 lines ;each line corresponds to the mean of each matrix

[Walter Roberson]

Using eval() is not recommended in this context.

[Azzi Abdelmalek]

why, M Walter? what is better?

[Oleg Komarov]

@Azzi: as you can see, the OP got here with this question because he used eval in the first place.

Also, your solution is not correct since you're overwriting result1.

The best solution (memory, time, legibility, safer etc...) in this case (when all results have the same dimension) is to use a 3D array.

[Azzi Abdelmalek]

M Komarov result1 is not overwritten, result1=sprintf('result(%d)',k) when k=2; result1='result(2)'. maby i didn't read all the question, I will remove one "mean'

[Oleg Komarov]

Ok, I agree. But, I think it is a very unhappy choice to call it result1 and assign it a dynamic string and then eval it.

Other methods are much clearer and less prone to errors.