[SciPy-User] How to average different pieces or an array?

[josef.pktd at gmail.com]

On Fri, Aug 7, 2009 at 1:38 PM, Borís BURLE<boris.burle at univ-provence.fr> wrote:
> Hi,
>
> Or another version, close to Gilles' one:
>
> x = array([1,3,2,6,7,4,5,4,9,4])
> y = [0,2,4,10]
> z = x
>
>
> for i in arange(len(y)-1):
>    z[y[i]:y[i+1]] = mean(x[y[i]:y[i+1]])
>
>
> hope this helps!
>
>     B.
>
> gilles Rochefort a écrit :
>
> Hi,
>
> Not sure to answer well to the question (indeed there is loops)  but
> have you tried
>  something like this :
>
> for s in [ slice(a,b) for a,b in zip(y[:-1],y[1:]) ]:
>     x[s] = mean(x[s])
>
> Regards,
> Gilles Rochefort.
>
>
>       I'm sure this is easy I just can't think of how to do it without
> a bunch of for loops which I would like to avoid since they take up so
> much computational time.
>
> Say I have an array x, with 100 entries.   Lets say I have another
> array y that looks like this:
>
> y = [0,5,10,20,40,100]  which specified which ranges of x I would like
> to average.
>
> In other words, I want elements 0 - 4  in x to be replaced by their
> average value. 5-9 replaced with their average value. 10-19 replaced
> by their averaged value, etc...
>
>     Effectively this will give me a new array with one 100 entries
> with different sized binning along the array.
>
>     If anyone knows how to do this without using for loops the whole
> way I would appreciate it.
>
>      Example:  I will make it simpler in case my above description was
> confusing.  Say x = [1,3,2,6,7,4,5,4,9,4] and I passed in y =
> [0,2,4,10].  The above description would return a new X:
>
> X = [2,2,4,4,5.5,5.5,5.5,5.5,5.5,5.5]
>
>     So that the apprpriate bins are averaged.
>
>                                        Joseph Smidt
>
>
>
> _______________________________________________
> SciPy-User mailing list
> SciPy-User at scipy.org
> http://mail.scipy.org/mailman/listinfo/scipy-user
>
>
>
> --
> Borís BURLE
> Laboratoire de Neurobiologie de la Cognition
> CNRS et Université de Provence
> tel: (+33) 4 88 57 68 79
> fax: (+33) 4 88 57 68 72
> web page:  http://www.up.univ-mrs.fr/lnc/ACT/act-fr.html
>
> _______________________________________________
> SciPy-User mailing list
> SciPy-User at scipy.org
> http://mail.scipy.org/mailman/listinfo/scipy-user
>
>

Using the loop of Boris to build the reusable label/indicator array
and bincount. labels may be slower because they don't assume that the
array is sorted by label.

>>> x = np.array([1,3,2,6,7,4,5,4,9,4])
>>> y = np.array([0,2,4,10])
>>> ind = np.empty(x.shape, int)
>>> for i in np.arange(len(y)-1): ind[y[i]:y[i+1]] = i

>>> means = np.bincount(ind,x)/np.bincount(ind)
>>> meanarr = means[ind]
>>> meanarr
array([ 2. ,  2. ,  4. ,  4. ,  5.5,  5.5,  5.5,  5.5,  5.5,  5.5])

Josef