ContinuousRV bug?

[F. B.]

I tried this:

In [1]: from sympy.stats import ContinuousRV, P

In [2]: var('x')
Out[2]: x

In [3]: X = ContinuousRV(x, exp(-x), (0, oo))

In [4]: P(X < Rational(1,2))

TypeError: Input must be Sets or iterables of Sets

I get the right answer when avoiding the stats module:

In [6]: var('k')
Out[6]: k

In [7]: cdf = integrate(exp(-x), (x, 0, k))

In [8]: solve(cdf - Rational(1, 2), k)
Out[8]: [log(2)]

[Abhishek K Das]

It requires an Interval as the third input parameter , what you are passing is basically a tuple of two numbers . If you write it as Interval(0,oo) instead of (0.oo) . 

It works fine ! 

[F. B.]

Sorry for the logic error in my previous post. In any case, what about this:

In [15]: X = ContinuousRV(x, exp(-x), Interval(0, oo))

In [16]: P(X < 1/2)
Out[16]:
   -1/2    
- ℯ     + 1

In [17]: P(X < log(2))
AttributeError: 'bool' object has no attribute 'is_Piecewise'

This one still looks like a bug.

[F. B.]

Sorry again, I wasn't using the latest version from master, in any case:

In [1]: from sympy.stats import *

In [2]: from sympy import *

In [3]: ;var x
Out[3]: x

In [4]: X = ContinuousRV(x, exp(-x), Interval(0, oo))

In [5]: P(X < log(2).n())
Out[5]: 0.500000000000000

In [6]: P(X < log(2))
Out[6]: oo

I would expect 1/2 as Out[6]

[Aaron Meurer]

That looks like an issue in the integration module. There should be
some way to get the integral that it computes, so that we can debug
from there.

Aaron Meurer

> --
> You received this message because you are subscribed to the Google Groups
> "sympy" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to sympy+un...@googlegroups.com.
> To post to this group, send email to sy...@googlegroups.com.
> Visit this group at http://groups.google.com/group/sympy.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/sympy/5024a59c-5ef1-480e-a1ce-e30b4c25d1f7%40googlegroups.com.
>
> For more options, visit https://groups.google.com/d/optout.

[Matthew Rocklin]

Generally when debugging sympy.stats I recommend inspecting the integrals that it produces using the evaluate=False keyword.  Sympy.stats almost always shoves off most of the hard work onto another SymPy module like integrals.

In [1]: from sympy.stats import *

In [2]: X = ContinuousRV(x, exp(-x), Interval(0, oo))

In [3]: P(X < log(2), evaluate=False)

Out[3]: 

0        

⌠        

⎮   -z   

⎮  ℯ     

⎮  ─── dz

⎮   2    

⌡        

-∞       

To view this discussion on the web visit https://groups.google.com/d/msgid/sympy/CAKgW%3D6%2BvrUtKCu1fU_78Hy7jYXLxWTE6V-YdOvrhBWtQ9T6_Cg%40mail.gmail.com.

[F. B.]

On Wednesday, May 14, 2014 11:28:54 PM UTC+2, Matthew wrote:

Generally when debugging sympy.stats I recommend inspecting the integrals that it produces using the evaluate=False keyword.  Sympy.stats almost always shoves off most of the hard work onto another SymPy module like integrals.

In [1]: from sympy.stats import *

In [2]: X = ContinuousRV(x, exp(-x), Interval(0, oo))

In [3]: P(X < log(2), evaluate=False)

Out[3]: 

0        

⌠        

⎮   -z   

⎮  ℯ     

⎮  ─── dz

⎮   2    

⌡        

-∞       

This function should be integrated from -log(2) to 0, not from -oo to 0.

I followed the debugger
 

> /home/.../sympy/sympy/stats/crv.py(392)compute_density()
    391         fy = sum(fx(g) * abs(g.diff(y)) for g in gs)
--> 392         return Lambda(y, fy)
    393

Line 391 of crv.py, it looks like the density function gets shifted by -log(2) without properly shifting the definition interval.

Previously, crv.where( ) failed to compute the proper interval with the log(2) condition, raising a NonImplementedError. Maybe it's just a bug in the interval transformation?