Splitapply using arrayfun or func with for loop

Hi,

I have a time series of observations. Each variable is in a separate vector. I need to calculate the following formula for each entity in my dataset:

for t = 1:n
  g(t+1) = g(t) * e(t+1) 
end

There are different number of observations for each entity.

To calculate the g(t+1) I tried to use the arrayfun and implement it in splitapply to apply this function to each entity. It looks somewhat like that:

a = @(g,e)arrayfun(@(i){g(i) .* e(i+1)},1:n-1));
result = splitapply(a,g,e,entity);

However, the following error occurs: "Error using splitapply (line 132), Index exceeds matrix dimensions." I also tried

result = splitapply(a(x,1),g,e,entity);

however here this occurs: "Index exceeds matrix dimensions.

Error in @(i){g(i).*e(i+1)}

Error in @(g,e)arrayfun(@(i){g(i).*e(i+1)},1:n-1))

So I think that this is just not the proper way to calculate this formula for each entity. Can someone please help ? What am I doing wrong?

Thanks in advance!

[Birdman]

If you have vectors which do not have the same size(different number of observations sentence means this), the dimension error will be inevitable.

[Guillaume]

arrayfun is definitively the wrong tool for a recursive equation as you have. You probably need to use filter instead.

However, it's really unclear what you are to trying to achieve overall with your splitapply. Can you provide a small example of input data and what you want as output.

Answer 1

[Andrei Bobrov]

out = accumarray(entity(:),1:numel(g),[],@(x){cumprod([g(x(1));e(x(2:end))],1)});
out = cell2mat(out);