Getting scipy.stats.models to work on Windows

classic Classic list List threaded Threaded
8 messages Options
Reply | Threaded
Open this post in threaded view
|

Getting scipy.stats.models to work on Windows

Wes McKinney
I'm trying to distance myself from using R for my research work, and
unfortunately I need to use Windows. I have been able to get the svn version of
scipy to work on my OS X machine, but on Windows the scipy.stats.models
regression models simply hang when I try to use them, ex, from test_rlm.py :

>>> from numpy.random import standard_normal as W
>>> X = W((40,10))
>>> import scipy.stats.models.rlm as rlm
>>> rlm(design=X)

--> Python freezes

Tried using stable (Numpy 1.0.4, scipy 0.6.0) releases and full latest SVN builds for both to no avail. Anyone have any luck with this?

Thanks,
Wes

_______________________________________________
SciPy-user mailing list
[hidden email]
http://projects.scipy.org/mailman/listinfo/scipy-user
Reply | Threaded
Open this post in threaded view
|

Re: Getting scipy.stats.models to work on Windows

Nils Wagner-2
On Mon, 7 Apr 2008 14:42:24 -0400
  "Wes McKinney" <[hidden email]> wrote:

> I'm trying to distance myself from using R for my
>research work, and
> unfortunately I need to use Windows. I have been able to
>get the svn version
> of
> scipy to work on my OS X machine, but on Windows the
>scipy.stats.models
> regression models simply hang when I try to use them,
>ex, from test_rlm.py :
>
>>>> from numpy.random import standard_normal as W
>>>> X = W((40,10))
>>>> import scipy.stats.models.rlm as rlm
>>>> rlm(design=X)
>
> --> Python freezes
>
> Tried using stable (Numpy 1.0.4, scipy 0.6.0) releases
>and full latest SVN
> builds for both to no avail. Anyone have any luck with
>this?
>
> Thanks,
> Wes

Works for me on linux.
 
>>> import scipy.stats.models.rlm as rlm
>>> rlm(design=X)
<scipy.stats.models.rlm.Model object at 0x41356fec>
>>> import numpy
>>> numpy.__version__
'1.0.5.dev4972'
>>> import scipy
>>> scipy.__version__
'0.7.0.dev4083'

Nils

_______________________________________________
SciPy-user mailing list
[hidden email]
http://projects.scipy.org/mailman/listinfo/scipy-user
Reply | Threaded
Open this post in threaded view
|

Re: Getting scipy.stats.models to work on Windows

Wes McKinney
Nils Wagner <nwagner <at> iam.uni-stuttgart.de> writes:

>
> Works for me on linux.
>

Works for me on OS X as well, but unfortunately not on Windows!

_______________________________________________
SciPy-user mailing list
[hidden email]
http://projects.scipy.org/mailman/listinfo/scipy-user
Reply | Threaded
Open this post in threaded view
|

Re: Getting scipy.stats.models to work on Windows

Robert Kern-2
In reply to this post by Wes McKinney
On Mon, Apr 7, 2008 at 11:42 AM, Wes McKinney <[hidden email]> wrote:

> I'm trying to distance myself from using R for my research work, and
> unfortunately I need to use Windows. I have been able to get the svn version
> of
> scipy to work on my OS X machine, but on Windows the scipy.stats.models
>  regression models simply hang when I try to use them, ex, from test_rlm.py
> :
>
> >>> from numpy.random import standard_normal as W
> >>> X = W((40,10))
> >>> import scipy.stats.models.rlm as rlm
>  >>> rlm(design=X)
>
> --> Python freezes
>
> Tried using stable (Numpy 1.0.4, scipy 0.6.0) releases and full latest SVN
> builds for both to no avail. Anyone have any luck with this?

Is your CPU capable of SSE2? Unfortunately, the official numpy binary
has been built with an ATLAS library that was compiled for a CPU
capable of SSE2. This means that the SSE2 instructions will cause
older CPUs to crash. The rlm() constructor does call some linear
algebra functions, so I suspect this is the problem.

There is a binary that was built without SSE2. When I dredge it up, I
will let you know.

--
Robert Kern

"I have come to believe that the whole world is an enigma, a harmless
enigma that is made terrible by our own mad attempt to interpret it as
though it had an underlying truth."
 -- Umberto Eco
_______________________________________________
SciPy-user mailing list
[hidden email]
http://projects.scipy.org/mailman/listinfo/scipy-user
Reply | Threaded
Open this post in threaded view
|

Re: Getting scipy.stats.models to work on Windows

Wes McKinney
On Mon, Apr 7, 2008 at 2:54 PM, Robert Kern <[hidden email]> wrote:


Is your CPU capable of SSE2? Unfortunately, the official numpy binary
has been built with an ATLAS library that was compiled for a CPU
capable of SSE2. This means that the SSE2 instructions will cause
older CPUs to crash. The rlm() constructor does call some linear
algebra functions, so I suspect this is the problem.

There is a binary that was built without SSE2. When I dredge it up, I
will let you know.

--
Robert Kern

"I have come to believe that the whole world is an enigma, a harmless
enigma that is made terrible by our own mad attempt to interpret it as
though it had an underlying truth."
 -- Umberto Eco
_______________________________________________
SciPy-user mailing list
[hidden email]
http://projects.scipy.org/mailman/listinfo/scipy-user


I'm running a fairly new machine, Pentium 4 dual-core, so seems perhaps unlikely. The machine passes the official NumPy and SciPy test suites.

- Wes

_______________________________________________
SciPy-user mailing list
[hidden email]
http://projects.scipy.org/mailman/listinfo/scipy-user
Reply | Threaded
Open this post in threaded view
|

Re: Getting scipy.stats.models to work on Windows

Alan G Isaac
In reply to this post by Robert Kern-2
On Mon, 7 Apr 2008, Robert Kern apparently wrote:
> There is a binary that was built without SSE2.

http://sourceforge.net/project/showfiles.php?group_id=1369&package_id=175103

http://sourceforge.net/project/showfiles.php?group_id=27747&package_id=19531&release_id=540981

Look for the binaries with "p3" in the name.

hth,
Alan Isaac



_______________________________________________
SciPy-user mailing list
[hidden email]
http://projects.scipy.org/mailman/listinfo/scipy-user
Reply | Threaded
Open this post in threaded view
|

Re: Getting scipy.stats.models to work on Windows

Wes McKinney


On Mon, Apr 7, 2008 at 3:04 PM, Alan G Isaac <[hidden email]> wrote:
On Mon, 7 Apr 2008, Robert Kern apparently wrote:
> There is a binary that was built without SSE2.

http://sourceforge.net/project/showfiles.php?group_id=1369&package_id=175103

http://sourceforge.net/project/showfiles.php?group_id=27747&package_id=19531&release_id=540981

Look for the binaries with "p3" in the name.

hth,
Alan Isaac



_______________________________________________
SciPy-user mailing list
[hidden email]
http://projects.scipy.org/mailman/listinfo/scipy-user


The problem apparently lies with numpy.linalg.pinv running on Windows, at least XP SP2 with Python 2.5.1, a quick search found that this has been a problem for more than just me. Anyone know anything about this?

>>> from numpy.random import standard_normal as W
>>> X = W((40,10))
>>> import numpy.linalg as L
>>> L.pinv(X)

frozen

_______________________________________________
SciPy-user mailing list
[hidden email]
http://projects.scipy.org/mailman/listinfo/scipy-user
Reply | Threaded
Open this post in threaded view
|

Re: Getting scipy.stats.models to work on Windows

Wes McKinney



The problem apparently lies with numpy.linalg.pinv running on Windows, at least XP SP2 with Python 2.5.1, a quick search found that this has been a problem for more than just me. Anyone know anything about this?


>>> from numpy.random import standard_normal as W
>>> X = W((40,10))
>>> import numpy.linalg as L
>>> L.pinv(X)

frozen


This page was helpful for anyone who should have this problem and find this thread:

http://www.scipy.org/scipy/numpy/ticket/627


_______________________________________________
SciPy-user mailing list
[hidden email]
http://projects.scipy.org/mailman/listinfo/scipy-user