Date: Mar 4, 2013 5:11 PM
Author: Derek Goring
Subject: Re: Best way to "classify" a vector? ("averaging" subsampling)
On Tuesday, March 5, 2013 10:19:09 AM UTC+13, Luca wrote:

> Hi everyone.

>

> Let's suppose to have these two vectors:

>

>

>

> t=0:0.01:1000; %timestamps

>

> val = rand(size(t)); %value associated with each timestamp

>

>

>

> Now let's suppose I want to calculate the mean value of val in each interval of "t" broad "1".

>

> What would be the best way to do it?

>

> The dumbest would be something _like_ (I'm sure somebody will find a couple of minor errors :-). But try to get the point!):

>

>

>

> t2 = ceil (t);

>

> [t2,I]= unique (t2,'first');

>

> [t2,J]=unique (t2,'last');

>

> meanVal= zeros(max(t2(:),1);

>

> for i=1:numel(meanVal)

>

> meanVal(i) = mean(val(I(i):val(J(i));

>

> end

>

>

>

> Now... I know that matlab always has some built in functions that does all of these basic operations better and faster.

>

> So.... what would be a smarter way to do this??

So, you're doing the equivalent of taking daily means (if the t were Matlab days).

Why is that dumb?

The only thing I'd criticise about your method is the use of i as an index. By default i and j are sqrt(-1). This may seem trivial, but one day this will jump up and bite you. Use something else, like it or jt, for the index.