This will not work.
Since the problem was fixed in the R source, what about just using it ?
# ---
import urllib
page =
urllib.urlopen("https://svn.r-project.org/R/branches/R-2-10-branch/src/library/stats/R/chisq.test.R";)
code = page.readlines()
chisq_test = ro.r(code)
# this will now work
# ( although not the most efficient way if larger vectors and/or a
# a large number of tests are made because of deparse(substitute()) )
chisq_test(ro.FloatVector(range(100)), ro.FloatVector(range(100)))
L.
Hao Fan wrote:
> Hi, Laurent and list,
>
> Thanks a lot for all the informations! I read through your
> discussions at
> https://stat.ethz.ch/pipermail/r-devel/2009-November/055701.html So it
> is a R problem not rpy. :-)
> I tried to work around this problem using the following code:
> -----------------------------------------------------------
> import rpy2.robjects as robjects
> import math
>
> p1dist = robjects.r.c()
> p2dist = robjects.r.c()
>
> num = 30
> for i in range(num):
> p1dist = robjects.r.c(p1dist, i)
> p2dist = robjects.r.c(p2dist, i+1)
>
> res = robjects.r['chisq.test'](p1dist, p2dist)
> print(str(res.r['p.value']).split()[-1])
> ----------------------------------------------------------
>
> It works well. However if the input values are not integer but float, it
> gives similar errors.
> -------------------------------------------------------------
> import rpy2.robjects as robjects
> import math
>
> p1dist = robjects.r.c()
> p2dist = robjects.r.c()
>
> num = 30
> for i in range(num):
> tmp1 = i * 0.1 + 0.01
> tmp2 = i * 0.1 + 0.05
> p1dist = robjects.r.c(p1dist, tmp1)
> p2dist = robjects.r.c(p2dist, tmp2)
>
> res = robjects.r['chisq.test'](p1dist, p2dist)
> print(str(res.r['p.value']).split()[-1])
> -------------------------------------------------------
> File "/usr/lib64/python2.6/site-packages/rpy2/robjects/__init__.py",
> line 422, in __call__
> res = super(RFunction, self).__call__(*new_args, **new_kwargs)
> rinterface.RRuntimeError: Error in names(dimnames(x)) <- DNAME :
> 'names' attribute [6] must be the same length as the vector [2]
>
> I would like to try what you suggested - bind the values to a symbol in
> an environment. Could you please point me to an example? I didn't find
> such a case in the R or rpy manual.
> (there is sth like the follows, but I tried sth similar and not work)
>
>>>> robjects.r('pi') + 2
> c(3.14159265358979, 2)
>>>> robjects.r('pi')[0] + 2
> 5.1415926535897931
>
> Thanks!
>
>
> Hao
>
>> This is odd.
>> When looking at what is happening, the problem is likely rooted in
>> the use of deparse(substitute(x)) and deparse(substitute(y)) in the code
>> for chisq.test.
>>
>> This is what is happening:
>>
>> >>> f = robjects.r('''function(x) return(deparse(substitute(x)))''')
>> >>> tuple(f(robjects.FloatVector(range(17))))
>> ('c(0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16)',)
>> >>> tuple(f(robjects.FloatVector(range(18))))
>> ('c(0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17', ')')
>>
>> deparse(substitute()) starts splitting the output into several elements.
>>
>> R allows the use of anonymous objects, but when created at the C level
>> some functions appear to break. I am currently seeing this as a
>> potential problem with R.
>>
>> One workaround is to bind your objects to a symbol in a given
>> environment, and call the chisq.test on them.
>> Not so elegant, but I do not think of an other temporary solution.
>>
>>
>> L.
>>
>>
>>
>> Hao Fan wrote:
>>> Hi, List and Laurent,
>>>
>>> For the chisq function I tried to used through rpy2, I just did
>>> more test with the exact code as follows:
>>> ----------------------------------------------
>>> import rpy2.robjects as robjects
>>>
>>> p1dist = []
>>> p2dist = []
>>> num = 17
>>> for x in range(num):
>>> p1dist.append(x)
>>> p2dist.append(x-1)
>>>
>>> print p1dist
>>> print p2dist
>>> rp1dist = robjects.FloatVector(p1dist)
>>> rp2dist = robjects.FloatVector(p2dist)
>>>
>>> res = robjects.r['chisq.test'](rp1dist, rp2dist)
>>> print(str(res.r['p.value']).split()[-1])
>>> -----------------------------------------------------
>>>
>>> When I set num to 17, it works
>>> "Warning message: In function (x, y = NULL, correct = TRUE, p =
>>> rep(1/length(x), length(x)), :
>>> Chi-squared approximation may be incorrect
>>> 0.2350833"
>>>
>>> However, if I set num to 18 (add one more element to the two vectors
>>> to be compared by chisq.test), it doesn't work.
>>> The error message is similar to my previous post:
>>> File "/usr/lib64/python2.6/site-packages/rpy2/robjects/__init__.py",
>>> line 422, in __call__
>>> res = super(RFunction, self).__call__(*new_args, **new_kwargs)
>>> rinterface.RRuntimeError: Error in names(dimnames(x)) <- DNAME :
>>> 'names' attribute [4] must be the same length as the vector [2]
>>>
>>> Could this mean that rpy2 doesn't take a long vector or I made some
>>> stupid mistake?
>>>
>>> Thanks!
>>>
>>>
>>> Hao
>>>> Hi Hao,
>>>>
>>>> The exact example that triggers the error may matter.
>>>>
>>>> I just tried the following with rpy2-2.1dev and it worked.
>>>>
>>>> x = robjects.FloatVector((1,2,3))
>>>> y = robjects.FloatVector((2,3,4))
>>>>
>>>> res = robjects.r['chisq.test'](x, y)
>>>>
>>>> I only get the following
>>>> Warning message:
>>>> In function (x, y = NULL, correct = TRUE, p = rep(1/length(x),
>>>> length(x)), :
>>>> Chi-squared approximation may be incorrect
>>>>
>>>> L.
>>>>
>>>>
>>>>
>>>> Hao Fan wrote:
>>>>> Hi, list
>>>>>
>>>>> I would like to calculate the fitness of two histograms, so I
>>>>> tried to use the chisq.test() in R through rpy2.
>>>>> The python code I have is as follows:
>>>>> -------------------
>>>>> import rpy2.robjects as robjects
>>>>>
>>>>> p1dist = [X1, X2, ... Xm]
>>>>> p2dist = [Y1, Y2, ... Ym]
>>>>>
>>>>> rp1dist = robjects.FloatVector(p1dist)
>>>>> rp2dist = robjects.FloatVector(p2dist)
>>>>>
>>>>> robjects.r['chisq.test'](rp1dist, rp2dist)
>>>>> -------------------
>>>>>
>>>>> Then I got the following error:
>>>>> File
>>>>> "/usr/lib64/python2.6/site-packages/rpy2/robjects/__init__.py", in
>>>>> __call__
>>>>> res = super(RFunction, self).__call__(*new_args, **new_kwargs)
>>>>> rinterface.RRuntimeError: Error in names(dimnames(x)) <- DNAME :
>>>>> 'names' attribute [62] must be the same length as the vector [2]
>>>>>
>>>>>
>>>>> I also tried rpy_classic, so my code change to the follows:
>>>>> --------------------
>>>>> import rpy2.robjects as robjects
>>>>> import rpy2.rpy_classic as rpy
>>>>> rpy.set_default_mode(rpy.NO_CONVERSION)
>>>>>
>>>>> p1dist = [X1, X2, ... Xm]
>>>>> p2dist = [Y1, Y2, ... Ym]
>>>>>
>>>>> rp1dist = robjects.FloatVector(p1dist)
>>>>> rp2dist = robjects.FloatVector(p2dist)
>>>>>
>>>>> robjects.r['chisq.test'](rp1dist, rp2dist)
>>>>> -----------------
>>>>>
>>>>> Then I got similar error:
>>>>> File "/usr/lib64/python2.6/site-packages/rpy2/rpy_classic.py", line
>>>>> 214, in __call__
>>>>> res = self.__sexp(*args_r, **kwargs_r)
>>>>> rinterface.RRuntimeError: Error in names(dimnames(x)) <- DNAME :
>>>>> 'names' attribute [62] must be the same length as the vector [2]
>>>>>
>>>>>
>>>>> I feel this problem is due to my wrong use of the R vector, but
>>>>> I can not find it. Any hints will be mostly appreciated!
>>>>> Thanks a lot in advance!
>>>>>
>>>>>
>>>>> Hao
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>
>>>
>
>
------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now. http://p.sf.net/sfu/bobj-july
_______________________________________________
rpy-list mailing list
rpy-list@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/rpy-list
