Mailman 3 September 2009 - SciPy-Dev

updating statsmodels license
by josef.pktd＠gmail.com Sept. 20, 2009

Sept. 20, 2009

A formality/legal question I want to update the official license file of scikits.statsmodels. Is the following header for the BSD license ok? partially infered from scipy svn history Thanks, Josef Copyright (C) 2006, Jonathan E. Taylor All rights reserved. Copyright (c) 2006-2008 Scipy Developers. All rights reserved. Copyright (c) 2009 Statsmodels Developers. All rights reserved.

1 0

scipy.stats: sf for symmetric distributions.
by David Warde-Farley Sept. 19, 2009

Sept. 19, 2009

Hi, I just noticed that the way _sf() is implemented in the base class is particularly horrible for continuous distributions that are symmetric about the origin, where 1.0 - cdf(x) is prone to roundoff error and could just as easily be calculated as cdf(-x). I wonder if this is a common enough case that that there ought to be an rv_continuous_zerosym class that candidate distributions inherit from, or if we should just override the method on a case by case basis. David

5 17

scipy.stats.chisquare insists second parameter is a numpy array
by Jason Grout Sept. 18, 2009

Sept. 18, 2009

In the code for scipy.stats.chisquare, the first parameter is automatically converted to a numpy array, but the second parameter isn't; it is just used as if it were a numpy array. Can the code be changed so that it does: f_obs = asarray(f_obs) k = len(f_obs) if f_exp is None: f_exp = array([np.sum(f_obs,axis=0)/float(k)] * len(f_obs),float) else: f_exp = asarray(f_exp,float) chisq = np.add.reduce((f_obs-f_exp)**2 / f_exp) return chisq, chisqprob(chisq, k-1) This will allow the user to specify a normal python list for f_exp, I believe. Thanks, Jason -- Jason Grout

1 0

fortran compiler arguments for scipy build?
by Hugo Shi Sept. 17, 2009

Sept. 17, 2009

I'm trying to build a 32 bit version of scipy on a 64 bit RHEL5 box. My build dies on fftpack, and it looks like it's cause gfortran needs to be called with the -m32 argument to make it compile for 32 bit systems. I've tried using config_fc, and distutils.cfg, and setup.cfg to make it do so, but nothing seems to work, can someone help me? I've tried doing this inthe command line - python setup.py config_fc --opt=m32 build as well as adding the following lines to configs [config_fc] opt=m32 in a local setup.cfg and in distutils.cfg in python2.6/distutil nothing seems to work. I've built numpy, lapack, BLAS and python all with a 32 bit target. Thanks in advanced.

1 0

scikits.statsmodels, new source distribution
by josef.pktd＠gmail.com Sept. 17, 2009

Sept. 17, 2009

This week, Skipper and I fixed some bugs, did some cleaning (renaming for consistency) and added more examples and additional code to the sandbox. I added a source distribution file at http://code.launchpad.net/statsmodels/trunk/initial-release/+download/sciki… which is the same as current trunk, but maybe more convenient for those without bazaar. This file doesn't contain the html docs since there were no changes since b1. We still have a few more improvements in mind before we will get to a release of scikits.statsmodels-0.1.0b2, but for those who want to try some regression in python with the webinar, the trunk is an improvement compared to b1 and passes the test suite. We welcome any bug reports and any ideas for improvements (and any comments, contributions, criticism,...) Josef

1 0

scipy.special.hyp2f1 bug?
by Skipper Seabold Sept. 17, 2009

Sept. 17, 2009

Since y'all are looking at the distributions vs. mpmath, it reminded me of this. I haven't filed a bug report yet, but I don't think one existed when I found this. I was working with the connection between the incomplete beta function and the hypergeometric2f1 function in scipy.special for GLM in statsmodels and I came across this blog post on mpmath about some corner cases for the hypergeometric2f1. <http://fredrik-j.blogspot.com/2009/06/hypergeometric-2f1-incomplete-beta.ht…> Two examples of discrepancies. In [1]: import mpmath In [2]: from scipy import special In [3]: mpmath.hyp2f1(3,-1,-1,0.5) Out[3]: mpf('2.5') In [4]: special.hyp2f1(3,-1,-1,.5) Out[4]: 8.0 In [5]: mpmath.hyp2f1(1,1,4,3+4j) Out[5]: mpc(real='0.49234384000963544', imag='0.60513406166123973') In [6]: special.hyp2f1(1,1,4,3+4j) Out[6]: (0.65925460474847919+1.3277887264473263j) I will file a bug report shortly, if we think mpmath is correct vs. scipy. -Skipper

4 5

SciPy SVN compilation on SnowLeopard
by Pierre GM Sept. 16, 2009

Sept. 16, 2009

All, I'm banging my head on some configuration issues for Scipy on Mac OS 10.6. I followed the recommendations from this site [http://blog.hyperjeff.net/?p=160 ], but there's obviously something very wrong in my system: _csr, spline, numpyio cannot be loaded because of a wrong architecture. (for example, the SciPy test suite gives me ImportError: dlopen(/Users/ pierregm/Computing/workspace/scipy-dev/scipy/io/numpyio.so, 2): no suitable image found. Did find: /Users/pierregm/Computing/workspace/ scipy-dev/scipy/io/numpyio.so: mach-o, but wrong architecture) Any hint would be welcome. And as a side note, I have to confirm bug #1194 of numpy (I have to symlink libnpymath.a to somewhere it could be found). Here are some relevant details gcc --version : i686-apple-darwin10-gcc-4.2.1 (GCC) 4.2.1 (Apple Inc. build 5646) (from Xcode 3.2, 64b) gfortran --version : GNU Fortran (GCC) 4.2.3 (from gfortran -m64 -v : Using built-in specs. Target: i686-apple-darwin8 Configured with: /Builds/unix/gcc/gcc-4.2/configure --prefix=/usr/ local --mandir=/share/man --program-transform-name=/^[cg][^.-]*$/s/ $/-4.2/ --build=i686-apple-darwin8 --host=i686-apple-darwin8 -- target=i686-apple-darwin8 --enable-languages=fortran Thread model: posix gcc version 4.2.3 F77=“gfortran -m64” FC=“gfortran -m64” buildlog: http://pastebin.com/m11b9627d

4 8

Time for a reality check
by David Goldsmith Sept. 15, 2009

Sept. 15, 2009

Hi, folks! I'm having a heIluva time trying to get pydocweb to run without "crashing" on my Kubuntu 9.0.4 install with Python2.6 and Django 1.0.2-1: I can runserver, view the front page, create and login to an admin account, view the control panel, enter and make changes in the admin site, but when I try to Pull from sources, I get either a slough of permission denied errors (if I didn't sudo manage.py) or (if I do sudo manage.py, which Pauli has since told me I should not have to do) a DjangoUnicodeDecodeError: 'ascii' codec can't decode byte 0xe2 in position 7504: ordinal not in range(128). You passed in RuntimeError(<string>) where <string> is longer than my Terminal's line retention buffer (which is 1000 lines). Pauli has me delving ever deeper into the pull scripts, but can't confirm that there isn't something in pydocweb that makes it incompatible w/ Python2.6, and is too busy w/ his dissertation (which I certainly understand) to help me substantially further. So, before I agonize any more, I'm hoping to find out: 0) if _anyone_ is presently successfully running (including the Pull from sources button/command) pydocweb under Python2.6 and Django 1.0.2-1; and 1) even if someone(s) is (are), if someone else running Python2.6 might be willing to take the time to try to install and run pydocweb (and all it's prereqs., of course) and let me know if they can successfully Pull from sources. Thanks! DG

2 1

API mismatch woes
by Pierre GM Sept. 14, 2009

Sept. 14, 2009

All, I like to use python -c "import setuptools; execfile('setup.py')" develop to install numpy in a virtual environment, which simplifies testing local modifications. Unfortunately, I keep running into the API mismatch problem, even after getting rid of the build directory and the *.pyc files. I'm sure I must be missing some simple step, but which one ? FYI, OS : Mac 10.6 python : Python 2.6.2 (r262:71600, Sep 11 2009, 18:38:58) (installed in its own /usr/local/Cellar) gcc : i686-apple-darwin10-gcc-4.2.1 (GCC) 4.2.1 Note that the following command works like a charm. pip install -E whatever -e svn+http://svn.scipy.org/svn/numpy/trunk/#egg=numpy

4 9

K2C accepts negative valued input
by David Goldsmith Sept. 13, 2009

Sept. 13, 2009

Is the following behavior: >>> from scipy.constants.constants import K2C >>> K2C(-40) -313.14999999999998 a bug or a feature? (I know I've read about sub-absolute-zero temperature physics, e.g., http://cryo.gsfc.nasa.gov/introduction/neg_Kelvin.html, but, quoting that site: "Physicists have defined a negative absolute temperature, but it's a bit flakey...") DG

2 3