[SciPy-user] step function

Jonathan Taylor jonathan.taylor at stanford.edu
Mon Mar 13 16:12:54 CST 2006


believe it or not, i just discovered "searchsorted"... my earlier hack 
was brutal, using "bisect"....

this is my current version and it seems to work fine.

class StepFunction:
    '''A basic step function: values at the ends are handled in the 
simplest way possible: everything to the left of x[0] is set to ival; 
everything to the right of x[-1] is set to y[-1].
   
    >>>
    >>> from numpy import *
    >>>
    >>> x = arange(20)
    >>> y = arange(20)
    >>>
    >>> f = StepFunction(x, y)
    >>>
    >>> print f(3.2)
    3
    >>> print f([[3.2,4.5],[24,-3.1]])
    [[ 3  4]
     [19  0]]
    >>>

    '''

    def __init__(self, x, y, ival=0., sorted=False):
           
        _x = N.array(x)
        _y = N.array(y)

        if _x.shape != _y.shape:
            raise ValueError, 'in StepFunction: x and y do not have the 
same shape'
        if len(_x.shape) != 1:
            raise ValueError, 'in StepFunction: x and y must be 
1-dimensional'

        self.x = N.array([-N.inf] + list(x))
        self.y = N.array([ival] + list(y))

        if not sorted:
            asort = N.argsort(self.x)
            self.x = N.take(self.x, asort)
            self.y = N.take(self.y, asort)
        self.n = self.x.shape[0]

    def __call__(self, time):

        tind = scipy.searchsorted(self.x, time) - 1
        _shape = tind.shape
        if tind.shape is ():
            return self.y[tind]
        else:
            tmp = N.take(self.y, tind)
            tmp.shape = _shape
            return tmp



Robert Kern wrote:

>Jonathan Taylor wrote:
>  
>
>>i was just wondering if there is a simple way to define a step function
>>in scipy, like an ECDF (empirical cumulative distribution function) in 
>>statistics....
>>
>>i have hacked my own version, but it is quite slow....
>>    
>>
>
>scipy.interpolate.interp1d has an option for nearest-neighbor interpolation
>which is almost what you want except that the provided data points are in the
>center of the steps instead of at the left or right.
>
>What does your version look like? Let's see if we can't improve upon it.
>
>  
>

-- 
------------------------------------------------------------------------
I'm part of the Team in Training: please support our efforts for the
Leukemia and Lymphoma Society!

http://www.active.com/donate/tntsvmb/tntsvmbJTaylor

GO TEAM !!!

------------------------------------------------------------------------
Jonathan Taylor                           Tel:   650.723.9230
Dept. of Statistics                       Fax:   650.725.8977
Sequoia Hall, 137                         www-stat.stanford.edu/~jtaylo
390 Serra Mall
Stanford, CA 94305



More information about the SciPy-user mailing list