| Re: MODE in IDL? [message #47103] |
Tue, 24 January 2006 19:20  |
David Fanning
Messages: 11724
Registered: August 2001
|
Senior Member |
|
|
Jonathan Greenberg writes:
> That is the defintion of mode -- it really surprised me this isn't built
> into IDL, since its a common and basic enough statistic.
Maybe it is and I've just never run into it. I relied on
you to do the literature search. :-)
Cheers,
David
--
David Fanning, Ph.D.
Fanning Software Consulting, Inc.
Coyote's Guide to IDL Programming: http://www.dfanning.com/
|
|
|
|
| Re: MODE in IDL? [message #47104 is a reply to message #47103] |
Tue, 24 January 2006 19:13  |
Jonathan Greenberg
Messages: 91
Registered: November 2002
|
Member |
|
|
David:
That is the defintion of mode -- it really surprised me this isn't built
into IDL, since its a common and basic enough statistic.
--j
"David Fanning" <davidf@dfanning.com> wrote in message
news:MPG.1e40988f4505ca03989b37@news.frii.com...
> Jonathan Greenberg writes:
>
>> I can't seem to find the function to calculate the mode of an array --
>> how
>> do I do this in IDL?
>
> I'm trying to remember back to grade school or somewhere, but
> isn't the mode the maximum of the frequency distribution. I
> remember that the mode isn't unique, because two numbers
> could have the same frequency in the sample (bimodal, I guess)
> and it is theoretically possible to have no mode (all frequencies
> are the same).
>
> But saying all that, if you have an integer array, I would
> think the mode is calculated like this:
>
> array = [1, 1, 2 , 4, 1, 3, 3, 2, 4, 5, 3, 2, 2, 1, 2, 6]
> h = Histogram(array, MIN=Min(array))
> bigfreq = Max(h)
> mode = Where(h EQ bigfreq) + Min(array)
> Print, mode
> 2
>
> Cheers,
>
> David
>
> --
> David Fanning, Ph.D.
> Fanning Software Consulting, Inc.
> Coyote's Guide to IDL Programming: http://www.dfanning.com/
|
|
|
|
| Re: MODE in IDL? [message #47105 is a reply to message #47104] |
Tue, 24 January 2006 19:04 |
David Fanning
Messages: 11724
Registered: August 2001
|
Senior Member |
|
|
Jonathan Greenberg writes:
> I can't seem to find the function to calculate the mode of an array -- how
> do I do this in IDL?
I'm trying to remember back to grade school or somewhere, but
isn't the mode the maximum of the frequency distribution. I
remember that the mode isn't unique, because two numbers
could have the same frequency in the sample (bimodal, I guess)
and it is theoretically possible to have no mode (all frequencies
are the same).
But saying all that, if you have an integer array, I would
think the mode is calculated like this:
array = [1, 1, 2 , 4, 1, 3, 3, 2, 4, 5, 3, 2, 2, 1, 2, 6]
h = Histogram(array, MIN=Min(array))
bigfreq = Max(h)
mode = Where(h EQ bigfreq) + Min(array)
Print, mode
2
Cheers,
David
--
David Fanning, Ph.D.
Fanning Software Consulting, Inc.
Coyote's Guide to IDL Programming: http://www.dfanning.com/
|
|
|
|
| Re: MODE in IDL? [message #47192 is a reply to message #47105] |
Wed, 25 January 2006 09:14 |
JD Smith
Messages: 850
Registered: December 1999
|
Senior Member |
|
|
On Tue, 24 Jan 2006 20:04:13 -0700, David Fanning wrote:
> array = [1, 1, 2 , 4, 1, 3, 3, 2, 4, 5, 3, 2, 2, 1, 2, 6]
> h = Histogram(array, MIN=Min(array))
> bigfreq = Max(h)
> mode = Where(h EQ bigfreq) + Min(array)
> Print, mode
> 2
Just a hint on your HISTOGRAM usage... this might be slightly preferred,
since it skips the WHERE and MIN:
array = [1, 1, 2 , 4, 1, 3, 3, 2, 4, 5, 3, 2, 2, 1, 2, 6]
void=max(histogram(array,OMIN=mn),mxpos)
mode=mn+mxpos
This method of course will be *very* problematic if you have, e.g.:
array = [1, 1, 2 , 4, 1, 3, 3, 2, 4, 5, 3, 2, 2, 1, 2, 6, 10000000]
Another option, if you worry about this, would be to re-cast as a
sorting problem, using the method discussed in a recent thread:
array=array[sort(array)]
wh=where(array ne shift(array,-1),cnt)
if cnt eq 0 then mode=array[0] else begin
void=max(wh-[-1,wh],mxpos)
mode=array[wh[mxpos]]
endelse
print,mode
Both methods will give you the lowest number in the case of ties for
the mode. The second will be slower, but more robust against large
dynamic range in your array. You could use both, deciding which to
use by the min/max of the array.
JD
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
