| Re: array manipulation (TOTAL-ing or MEDIAN-ing) in uneven bins [message #82472 is a reply to message #82403] |
Thu, 13 December 2012 12:11   |
havok2063
Messages: 24
Registered: December 2012
|
Junior Member |
|
|
On Wednesday, December 12, 2012 5:19:07 PM UTC-5, Jeremy Bailin wrote:
> On 12/12/12 4:18 PM, Jeremy Bailin wrote:
>
>> On 12/12/12 4:03 PM, Jeremy Bailin wrote:
>
>>> On 12/12/12 10:16 AM, havok2063@gmail.com wrote:
>
>>>>
>
>>>> I have several unrelated problems that I'm solving in the same
>
>>>> efficient way (with loops). I'm trying to perform some array
>
>>>> operation on an array, according to a list of (let's call them) uneven
>
>>>> bins.
>
>>>>
>
>>>> I have an array, say d, of 146 elements. I have a separate array that
>
>>>> represents uneven bins that I want to perform the operation on, like
>
>>>> MEDIAN, or TOTAL. For example,
>
>>>>
>
>>>> ntot = [15,45,56,90,116,146]
>
>>>>
>
>>>> I want as output an array, of 6 elements, that contains the MEDIAN (or
>
>>>> TOTAL) of array d according to the indices listed in ntot.
>
>>>>
>
>>>> So the 1st element would contain median(d[0:14],/even), the 2nd
>
>>>> median(d[15:44],/even), etc....
>
>>>>
>
>>>> Or the same thing with total....total(d[0:14]), total(d[15:44]) , etc...
>
>>>>
>
>>>> Right now I'm looping over the number of elements in ntot to do this
>
>>>> and I don't much care for loops.
>
>>>>
>
>>>> I don't think this is quite the same thing as the example given in the
>
>>>> "Horror and Disgust of Histogram" article nor does this sound like
>
>>>> something I can do with value_locate, although I'm not too familiar
>
>>>> with value_locate.
>
>>>>
>
>>>> Any ideas on this? Thanks a lot.
>
>>>>
>
>>>
>
>>> As David says, this screams VALUE_LOCATE. And HISTOGRAM. They play very
>
>>> nicely together for this sort of problem!
>
>>>
>
>>> First we need to label the bin for each element:
>
>>>
>
>>> nelements = 146
>
>>> binlabel = value_locate(ntot, lindgen(nelements))
>
>>>
>
>>> Then use histogram to group the elements by bin label. Notice that the
>
>>> way you've defined ntot, elements 0 through 14 will be labelled "-1" by
>
>>> value_locate, so we start the histogram there:
>
>>>
>
>>> nbin = n_elements(ntot)
>
>>> hist = histogram(binlabel, min=-1, max=nbin-1, reverse_indices=ri)
>
>>>
>
>>> And finally we do the usual loop through the reverse indices to
>
>>> calculate the statistics:
>
>>>
>
>>> medianbin = fltarr(nbin)
>
>>> totbin = fltarr(nbin)
>
>>> for i=0L,nbin-1 do if hist[i] gt 0 then begin
>
>>> these = ri[ri[i]:ri[i+1]-1]
>
>>> medianbin[i] = median(d[these], /even)
>
>>> totbin[i] = total(d[these])
>
>>> endif
>
>>>
>
>>> -Jeremy.
>
>>
>
>> Actually, for the total you can do a lot better by using cumulative:
>
>>
>
>> runningtotal = total(d, /cumulative)
>
>> totbin = runningtotal[ntot] - [0,runningtotal[ntot]]
>
>>
>
>> -Jeremy.
>
>
>
> Ack, hit send to soon. That should be:
>
>
>
> runningtotal = total(d, /cumulative)
>
> totbin = runningtotal[ntot-1] - [0,runningtotal[ntot-1]]
>
>
>
> -Jeremy.
Excellent. That really hits the spot. Thanks a lot. I was hovering somewhere around there but couldn't quite converge on what to do with value_locate.
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
