judithabk6
commented
9 months ago I think I have fixed the first issue (see branch fix_conversion_r_matrix), but I have the second issue in the new version. There is some problem if you import the function from somewhere else than where rpy2 is imported. Do you have any idea @bthirion? I am really blocked.
here is the exact code
In [1]: from med_bench.get_simulated_data import simulate_data
...: from numpy.random import default_rng
...: import numpy as np
...: from med_bench.mediation import r_mediation_DML
In [2]: SAMPLE_SIZE = 1000
...: CONFIG = 1
...:
...: data = simulate_data(
...: n=SAMPLE_SIZE,
...: rg=default_rng(45),
...: mis_spec_m=False,
...: mis_spec_y=False,
...: dim_x=5,
...: dim_m=1,
...: seed=9,
...: type_m="binary",
...: sigma_y=0.5,
...: sigma_m=0.5,
...: beta_t_factor=0.5,
...: beta_m_factor=0.5,
...: )
...:
...: x, t, m, y = data[0], data[1], data[2], data[3]
...:
...: r_mediation_DML(x, t, m, y, trim=0.0, order=1)
R[write to console]: Error in glmnet::glmnet(x, y, family = c("binomial"), alpha = 1, lambda = lambda[1], :
x should be a matrix with 2 or more columns
---------------------------------------------------------------------------
RRuntimeError Traceback (most recent call last)
Cell In[2], line 21
4 data = simulate_data(
5 n=SAMPLE_SIZE,
6 rg=default_rng(45),
(...)
16 beta_m_factor=0.5,
17 )
19 x, t, m, y = data[0], data[1], data[2], data[3]
---> 21 r_mediation_DML(x, t, m, y, trim=0.0, order=1)
File /data/parietal/store/work/jabecass/med_bench/src/med_bench/mediation.py:962, in r_mediation_DML(y, t, m, x, trim, order)
930 """
931 y array-like, shape (n_samples)
932 outcome value for each unit, continuous
(...)
958 Powers command of the LARF package.
959 """
960 x_r, t_r, m_r, y_r = [_convert_array_to_R(uu) for uu in
961 (x, t, m, y)]
--> 962 res = causalweight.medDML(y_r, t_r, m_r, x_r, trim=trim, order=order)
963 raw_res_R = np.array(res.rx2('results'))
964 ntrimmed = res.rx2('ntrimmed')[0]
File /data/parietal/store/work/jabecass/miniconda3/envs/causal2024/lib/python3.10/site-packages/rpy2/robjects/functions.py:208, in SignatureTranslatedFunction.__call__(self, *args, **kwargs)
206 v = kwargs.pop(k)
207 kwargs[r_k] = v
--> 208 return (super(SignatureTranslatedFunction, self)
209 .__call__(*args, **kwargs))
File /data/parietal/store/work/jabecass/miniconda3/envs/causal2024/lib/python3.10/site-packages/rpy2/robjects/functions.py:131, in Function.__call__(self, *args, **kwargs)
129 else:
130 new_kwargs[k] = cv.py2rpy(v)
--> 131 res = super(Function, self).__call__(*new_args, **new_kwargs)
132 res = cv.rpy2py(res)
133 return res
File /data/parietal/store/work/jabecass/miniconda3/envs/causal2024/lib/python3.10/site-packages/rpy2/rinterface_lib/conversion.py:45, in _cdata_res_to_rinterface.<locals>._(*args, **kwargs)
44 def _(*args, **kwargs):
---> 45 cdata = function(*args, **kwargs)
46 # TODO: test cdata is of the expected CType
47 return _cdata_to_rinterface(cdata)
File /data/parietal/store/work/jabecass/miniconda3/envs/causal2024/lib/python3.10/site-packages/rpy2/rinterface.py:817, in SexpClosure.__call__(self, *args, **kwargs)
810 res = rmemory.protect(
811 openrlib.rlib.R_tryEval(
812 call_r,
813 call_context.__sexp__._cdata,
814 error_occured)
815 )
816 if error_occured[0]:
--> 817 raise embedded.RRuntimeError(_rinterface._geterrmessage())
818 return res
RRuntimeError: Error in glmnet::glmnet(x, y, family = c("binomial"), alpha = 1, lambda = lambda[1], :
x should be a matrix with 2 or more columnsbut then if you add
In [3]: from med_bench.utils.utils import get_interactions, _convert_array_to_R
In [4]: _convert_array_to_R(x)
Out[4]: <rpy2.rinterface.FloatSexpVector object at 0x7f920ce52840> [RTYPES.REALSXP]
In [5]: import rpy2.robjects.packages as rpackages
In [6]: causalweight = rpackages.importr('causalweight')
In [7]: x_r, t_r, m_r, y_r = [_convert_array_to_R(uu) for uu in (x, t, m, y)]
In [8]: res = causalweight.medDML(y_r, t_r, m_r, x_r, trim=0.0, order=1)
In [9]: it works
bthirion
I have 2 different issues The first one is a change of behavior of the utils function https://github.com/judithabk6/med_bench/blob/2cd241ccd6880baa4ebd815336d7bb8dc48363df/src/med_bench/utils/utils.py#L52
is different in my new environment (more recent rpy2)
and with the more recent rpy2
and this breaks some estimators of course.
And even in the env with the old rpy2, there is a weird behavior.
but executing directly what is in the function works
I think it's related to the new structure introduced by #58 (import has not the same behavior)
I propose to fix everything for the more recent rpy2 (the old rpy2 also breaks the import of
RRuntimeErrorfrom rpy2, which is why I changed in the first place)