How can I loop though a vector select certain elements and store them in a new vector?

I want to have a function that calculates the average of an input vector that satisfies certain conditions. Here is how I imagine my algorithm:

1. user inputs a vector ,r, and 2 scalars,u and l (upperbound and lowerbound)
2. script loops though r and evaluates each expression.
3. If u>r<l then select value and add to a new vector, v. (vectorize?)
4. Final output should be average=sum(v)/size(v)

My code so far:

function averageRate = fermentationRate(r, l, u)
[row, col]=size(r);    %dimensions of inputvec,r 
for i=1:col            %iterates through 1 to last column in the inputvec,r
if r>l && r<u          %condtions that must be true before value can be used
 v=0+(r);
 [row1,col1]=size(v)   %dimension of vector,v containing useable inputs
 averageRate=sum(v)/col1; %average of inputs 
end 
end
end

Error in Fermenationproblem (line 8)
[row, col]=size(r);    %dimensions of inputvec,r

>> averageRate([20.1 19.3 1.1 18.2 19.7 121.1 20.3 20.0], 15,25)
Undefined function or variable 'averageRate'.

Answer 1

[Cam Salzberger]

Hey Mark,

There is a very convenient feature in MATLAB called logical indexing. This can allow you to vectorize your code, and drastically improve efficiency. There is also the built-in mean function to determine average, rather than doing sum/count.

% Figure out which vector elements are within bounds
% Needs to use element-wise &, not logical &&
idx = r > l & r < u;
% Determine average:
averageRate = mean(r(idx));

Also, I'd recommend not using i, j, or l for variable names. i and j have built-in meanings in MATLAB (imaginary number), and l looks too much like 1.

-Cam

[mark dodds]

Thank you so much Cam Salzberger. It works flawless now. Could the problem be solved using loops? Or was I completely offtrack? Thanks again! -Mark

[Cam Salzberger]

No, you absolutely could use loops, it just wouldn't be very efficient. The issue is that you'd either have to define a second vector larger than you need, or build the vector as you go without preallocation.

If you wanted to do loops without running into the preallocation issue, it'd probably make sense to not bother building a second vector at all, and just process the values on the fly:

total = 0;
count = 0;
for k = numel(r)
    if r(k) > l && r(k) < u
        total = total + r(k);
        count = count + 1;
    end
end
averageRate = total/count;

[mark dodds]

Ah, okay that makes sense! Thanks a lot Cam!

    for k = 1: numel(r)

[Cam Salzberger]

Oops, yeah, my bad. Thanks for catching that.