Breaking and spliting vector into groups and averaging

Hi,

I have a long one-dimensional time vector that I want to split in chunks and average each chunk. I have another logical vector (same length) of 0s and 1s that changes the sign to signal each chunk transition. How could I easily use the information on the logical vector to select each chunk of the time vector?

Thanks!

timevec = [0 2 3 4 5 0 9 3 4];

chunk = [0 0 0 1 1 1 1 0 0];

Answer 1

[Adam Danz]

timevec = [0 2 3 4 5 0 9 3 4];
chunk = [0 0 0 1 1 1 1 0 0];

g = cumsum([1, diff(chunk)]~=0); % assuming 'chunk' is a row vector
groupMeans = splitapply(@mean, timevec, g)

groupMeans = 1×3

1.6667 4.5000 3.5000

[Albert]

Excellent! Fast and effective, thanks!

[Albert]

I had another question. How could I update the code above so that I group on every period? So instead of taking groups every change in the chunk vector, taking them every time it's back to 0. For example:

timevec = [1 2 3 4 5 6 7 8 9 10 11 12];

chunk = [0 0 0 1 1 1 1 0 0 0 0 1];

should do the means in two groups for [1 2 3 4 5 6 7] and for [8 9 10 11 12]

thanks!

[Albert]

And there is another point, what if I would like to apply a condition to the mean? Once the chunks are ready, in the splitapply part I would like to apply the mean of each chunk but only to the indexes fulfilling a certain condition expressed by another vector cond of same length...

thanks again!

timevec = [0 2 3 4 5 0 9 3 4];

chunk = [0 0 0 1 1 1 1 0 0];

cond = [1 1 1 1 0 0 1 1 1];

g = cumsum([1, diff(chunk)]~=0); % assuming 'chunk' is a row vector

groupMeans = splitapply(@mean, timevec, g)

[Adam Danz]

To define groups by the chunk vector resetting to 0,

chunk =   [0 0 0 1 1 1 1 0 0 0 0 1]

chunk = 1×12

0 0 0 1 1 1 1 0 0 0 0 1

g = cumsum(diff([1,chunk])==-1)

g = 1×12

1 1 1 1 1 1 1 2 2 2 2 2

For the second question, do you mean this?

timevec = [0 2 3 4 5 0 9 3 4];
chunk = [0 0 0 1 1 1 1 0 0];
cond =  logical([1 1 1 1 0 0 1 1 1]);
X = timevec(cond); 
Y = chunk(cond); 

g = cumsum([1, diff(Y)]~=0); 
groupMeans = splitapply(@mean, X, g)

[Albert]

For the first question, great, I didn't think it was so easy the adaptation!

For the second question it's a bit trickier because the Y array, after applying the condition in cond may totally mess around the chunks computation in the lines later on (transitions may be lost after applying the condition!). As a workaround I computed the chunks for all the values and applied the condition afterwards when computing the mean through a custom function as:

timevec = [0 2 3 4 5 0 9 3 4];
chunk = [0 0 0 1 1 1 1 0 0];
cond =  logical([1 1 1 1 0 0 1 1 1]);

g = cumsum([1, diff(chunk)]~=0); 
groupMeans = splitapply(@(vec,con)mean(vec(con)), timevec, cond, g)