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
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>
>>
--
Hao Fan, PhD
Department of Biopharmaceutical Sciences
Mission Bay, Byers Hall
1700 4th Street, Suite 501
University of California, San Francisco
San Francisco, CA 94158-2330
Tel +1 (415) 514-4258, 4232, 4233
Fax +1 (415) 514-4231, 4234
------------------------------------------------------------------------------
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
