Add more documentation

docs/reference.rst

@@ -8,6 +8,7 @@ Seldonian Algorithm
 .. automodule:: seldonian.algorithm
       :members:
       :show-inheritance:
+      :private-members: _safetyTest
 
 ==========================
 Seldonian Abstract classes
@@ -18,6 +19,7 @@ Use this as a base class to implement your own fair model using the Seldonian ap
       :members:
       :show-inheritance:
       :special-members:
+      :private-members: _safetyTest
 
 ===========================
 Sample constraint functions

seldonian/algorithm.py

@@ -36,25 +36,22 @@ def predict(self, X):
     @abstractmethod
     def _safetyTest(self, **kwargs):
         """
-        Run the safety test on the trained model from the candidate selection part i.e. the
-        :func:`fit` function.
-        It is also used to predict the :math:`g(\\theta)` value used in candidate selection.
+        Run the safety test on the trained model from the candidate selection part i.e. the :func:`fit` function. It is also used to predict the :math:`g(\\theta)` value used in candidate selection.
 
         :param kwargs Key value arguments sent to the subclass implementation of safety test.
         :return Depending on the implementation, it will either return `0` if it passes or
         `1` if it doesn't. Or, it will also return the :math:`g(\\theta)` value
-        if it does not pass the safety test.
-
-        Either way, use the ``safetyTest()`` method (with the _) to get a boolean value.
+        if it does not pass the safety test. Use the ``safetyTest()`` method to get a boolean value.
         """
         raise NotImplementedError("_safetyTest function must be implemented")
         pass
 
     def safetyTest(self, **kwargs):
         """
-        A wrapper for the ``_safetyTest`` method that return a ``Boolean`` indicating whther the
+        A wrapper for the ``_safetyTest`` method that return a ``Boolean`` indicating whether the
         model passed the safety test.
-        :param kwargs: Key-value arguments that is passed directly to ``_safetyTest`.
+
+        :param kwargs: Key-value arguments that is passed directly to ``_safetyTest``.
         :return:
         - ``True`` if model passed the safety test.
         - ``False`` if the model fails the safety test.

seldonian/seldonian.py

@@ -344,11 +344,12 @@ def data(self):
 
     def _safetyTest(self, theta=None, predict=False, ub=True):
         """
-        This
-        :param theta:
-        :param predict:
-        :param ub:
-        :return:
+        This is the mehtod that implements the safety test. for this model.
+
+        :param theta: Model parameters to be used to run the safety test. **Default** - ``None``. If ``None``, the current model parameters used.
+        :param predict: **Default** - ``False``. Indicate whether you want to predict the upper bound of :math:`g(\\theta)` using the candidate set (this is used when running candidate selection).
+        :param ub: returns the upper bound if ``True``. Else, it returns the calculated value. **Default**- ``True``.
+        :return: Returns the value :math:`max\{0, g(\\theta) | X\}` if `predict` = ``False`` ,  else :math:`max\{0, \\hat{g}(\\theta) | X\}`.
         """
         if theta is None:
             theta = self.theta