Merge pull request #111 from theislab/dev

Dev v0.6.11

Author: davidsebfischer

diffxpy/models/batch_bfgs/optim.py

@@ -6,7 +6,6 @@
 from numpy.linalg import pinv
 import numpy.random
 from multiprocessing import Pool
-import xarray as xr
 
 from .objectives import *
 
@@ -22,12 +21,12 @@ def __init__(self, Estim_BFGS, nproc):
         self._observations = Estim_BFGS.x.shape[0]
         self._design_loc = Estim_BFGS.design_loc
         self._design_scale = Estim_BFGS.design_scale
-        self._loss = xr.DataArray(Estim_BFGS.full_loss(nproc))
+        self._loss = Estim_BFGS.full_loss(nproc)
         self._log_probs = -self._loss
         self._probs = np.exp(self._log_probs)
-        self._mles = xr.DataArray(np.transpose(Estim_BFGS.mles()))
-        self._gradient = xr.DataArray(np.zeros([Estim_BFGS.x.shape[1]]))
-        self._fisher_inv = xr.DataArray(Estim_BFGS.fisher_inv)
+        self._mles = np.transpose(Estim_BFGS.mles())
+        self._gradient = np.zeros([Estim_BFGS.x.shape[1]])
+        self._fisher_inv = Estim_BFGS.fisher_inv
         self._idx_loc = np.arange(0, Estim_BFGS.design_loc.shape[1])
         self._idx_scale = np.arange(Estim_BFGS.design_loc.shape[1],
                                     Estim_BFGS.design_loc.shape[1] + Estim_BFGS.design_scale.shape[1])

diffxpy/models/hessian.py

@@ -27,9 +27,9 @@ def hes_nb_glm_mean_block(
         Estimated mean parameters across cells for a given gene.
     :param mu: np.ndarray (cells,)
         Estimated dispersion parameters across cells for a given gene.
-    :param design_loc: np.ndarray, matrix, xarray (cells, #parameters location model)
+    :param design_loc: np.ndarray, matrix (cells, #parameters location model)
         Design matrix of location model.
-    :param design_scale: np.ndarray, matrix, xarray (cells, #parameters shape model)
+    :param design_scale: np.ndarray, matrix (cells, #parameters shape model)
         Design matrix of shape model.
     :param i: int
         Index of first dimension in fisher information matrix which is to be computed.
@@ -78,9 +78,9 @@ def hes_nb_glm_disp_block(
         Estimated mean parameters across cells for a given gene.
     :param mu: np.ndarray (cells,)
         Estimated dispersion parameters across cells for a given gene.
-    :param design_loc: np.ndarray, matrix, xarray (cells, #parameters location model)
+    :param design_loc: np.ndarray, matrix (cells, #parameters location model)
         Design matrix of location model.
-    :param design_scale: np.ndarray, matrix, xarray (cells, #parameters shape model)
+    :param design_scale: np.ndarray, matrix (cells, #parameters shape model)
         Design matrix of shape model.
     :param i: int
         Index of first dimension in fisher information matrix which is to be computed.
@@ -129,9 +129,9 @@ def hes_nb_glm_meandisp_block(
         Estimated mean parameters across cells for a given gene.
     :param mu: np.ndarray (cells,)
         Estimated dispersion parameters across cells for a given gene.
-    :param design_loc: np.ndarray, matrix, xarray (cells, #parameters location model)
+    :param design_loc: np.ndarray, matrix (cells, #parameters location model)
         Design matrix of location model.
-    :param design_scale: np.ndarray, matrix, xarray (cells, #parameters shape model)
+    :param design_scale: np.ndarray, matrix (cells, #parameters shape model)
         Design matrix of shape model.
     :param i: int
         Index of first dimension in fisher information matrix which is to be computed.
@@ -160,9 +160,9 @@ def hes_nb_glm_bygene(
         Estimated mean parameters across cells for a given gene.
     :param mu: np.ndarray (cells,)
         Estimated dispersion parameters across cells for a given gene.
-    :param design_loc: np.ndarray, matrix, xarray (cells, #parameters location model)
+    :param design_loc: np.ndarray, matrix (cells, #parameters location model)
         Design matrix of location model.
-    :param design_scale: np.ndarray, matrix, xarray (cells, #parameters shape model)
+    :param design_scale: np.ndarray, matrix (cells, #parameters shape model)
         Design matrix of shape model.
     
     :return: np.ndarray (#parameters location model + #parameters shape model, #parameters location model + #parameters shape model)
@@ -212,9 +212,9 @@ def theta_covar_bygene(
         Estimated mean parameters across cells for a given gene.
     :param mu: np.ndarray (cells,)
         Estimated dispersion parameters across cells for a given gene.
-    :param design_loc: np.ndarray, matrix, xarray (cells, #parameters location model)
+    :param design_loc: np.ndarray, matrix (cells, #parameters location model)
         Design matrix of location model.
-    :param design_scale: np.ndarray, matrix, xarray (cells, #parameters shape model)
+    :param design_scale: np.ndarray, matrix(cells, #parameters shape model)
         Design matrix of shape model.
     
     :return: np.ndarray (#parameters location model + #parameters shape model, #parameters location model + #parameters shape model)
@@ -243,9 +243,9 @@ def theta_sd_bygene(
         Estimated mean parameters across cells for a given gene.
     :param disp: np.ndarray (cells,)
         Estimated dispersion parameters across cells for a given gene.
-    :param design_loc: np.ndarray, matrix, xarray (cells, #parameters location model)
+    :param design_loc: np.ndarray, matrix (cells, #parameters location model)
         Design matrix of location model.
-    :param design_scale: np.ndarray, matrix, xarray (cells, #parameters shape model)
+    :param design_scale: np.ndarray, matrix (cells, #parameters shape model)
         Design matrix of shape model.
     
     :return: np.ndarray (#parameters location model + #parameters shape model,)

diffxpy/testing/det.py

@@ -581,10 +581,8 @@ def log_fold_change(self, base=np.e, return_type="vector"):
 
             - "dataframe":
               return a pandas.DataFrame with columns `gene`, `minuend_<group>`, `subtrahend_<group>` and `logFC`.
-            - "xarray":
-              return a xarray.DataArray with dimensions `(minuend, subtrahend, gene)`
 
-        :return: either pandas.DataFrame or xarray.DataArray
+        :return:
         """
         factors = set(self.full_design_loc_info.term_names) - set(self.reduced_design_loc_info.term_names)
 

diffxpy/testing/tests.py

@@ -594,7 +594,7 @@ def wald(
                 # there is no such column; modify design matrix to create one
                 design_loc[:, col_indices] = np.where(samples, 1, 0)
     elif coef_to_test is not None:
-        # Directly select coefficients to test from design matrix (xarray):
+        # Directly select coefficients to test from design matrix:
         # Check that coefficients to test are not dependent parameters if constraints are given:
         coef_loc_names = glm.data.view_coef_names(design_loc).tolist()
         if not np.all([x in coef_loc_names for x in coef_to_test]):

diffxpy/testing/utils.py

@@ -123,7 +123,7 @@ def design_matrix(
         formula: Union[None, str] = None,
         as_numeric: Union[List[str], Tuple[str], str] = (),
         dmat: Union[pd.DataFrame, None] = None,
-        return_type: str = "xarray"
+        return_type: str = "patsy"
 ) -> Union[patsy.design_info.DesignMatrix, pd.DataFrame]:
     """ Create a design matrix from some sample description.
 

docs/requires.txt

@@ -2,7 +2,7 @@ numpy>=1.14.0
 scipy
 pandas
 patsy>=0.5.0
-batchglm>=0.6.5
+batchglm>=0.6.6
 statsmodels
 anndata
 seaborn

requirements.txt

@@ -2,7 +2,7 @@ numpy>=1.14.0
 scipy
 pandas
 patsy>=0.5.0
-batchglm>=0.6.5
+batchglm>=0.6.6
 statsmodels
 anndata
 seaborn

setup.py

@@ -21,7 +21,7 @@
         'scipy>=1.2.1',
         'pandas',
         'patsy>=0.5.0',
-        'batchglm>=0.6.5',
+        'batchglm>=0.6.6',
         'statsmodels',
     ],
     extras_require={