unequal arrays and basic calculations

I want to reshape vector num to an averages vector as

Day    time    num
20050103  34200000  10
20050103  34205000  14
20050103  34210000  8
20050103  34215000  11
20050103  34220000  15
20050103  34225000  13
20050103  34230000   7

Assume I want to average over time each 4 intervals

Then I should have something like: 10.75 and 11.67

10.75= (10+14+8+11)/4
11.67=(15+13+7)/3

If I do something like

av=mean(reshape(num, [4 length(num/4]))';

I obviously get an issue with the size of the vectors

Is there an easy fix to the problem?

Merry Xmas and thank you

Answer 1

[Junaid]

Dear I guess if your matrix num after reshaping maintains it dimensions then reshape function should not be a problem. The examples values you have given, i executed following code without any error.

num = [20050103
    34200000
          10
    20050103
    34205000
          14
    20050103
    34210000
           8
    20050103
    34215000
          11
    20050103
    34220000
          15
    20050103
    34225000
          13
    20050103
    34230000];
 av=reshape(num, 4,length(num)/4); % this is 4 x 5 matrix. 

m = mean(av); % col wise sum
mr = mean(av,2); % row wise sum.

I hope I understood your question ?

Answer 2

[Walter Roberson]

MATLAB cannot have columns of different lengths in a numeric matrix, and so reshape() can only be used when the initial number of elements matches the requested number of full rows times the requested number of full columns.

You can pad the original vector with a value that you then program to be ignored afterwards. For example, you could pad your vector with NaN and then use nanmean() from one of the toolboxes or from the user contribution http://www.mathworks.com/matlabcentral/fileexchange/6837

Answer 3

[Andrei Bobrov]

try this cod. Use function nanmean by Statistic Toolbox

num = [10 14 8 11 15 13 7].';

n = 4;
num1 = [num; NaN(n-rem(numel(num),n),1)];
out = nanmean(reshape(num1,n,[])).';

variant without use nanmean

n = 4;
n1 = numel(num);
n2 = ceil(n1/n);
num1 = NaN(n,n2);
num1(1:n1) = num;
out = arrayfun(@(x)mean(num1(~isnan(num1(:,x)),x),(1:n2)');