docs and capecod hotfix

MANIFEST.in

@@ -13,3 +13,4 @@ include chainladder/utils/data/raa.csv
 include chainladder/utils/data/ukmotor.csv
 include chainladder/utils/data/usaa.csv
 include chainladder/utils/data/usauto.csv
+include chainladder/utils/data/cc_sample.csv

chainladder/__init__.py

@@ -5,4 +5,4 @@
 from chainladder.methods import *
 from chainladder.workflow import *
 
-__version__ = '0.2.3'
+__version__ = '0.2.4'

chainladder/core/base.py

@@ -200,7 +200,24 @@ def cum_to_incr(self, inplace=False):
 
     @check_triangle_postcondition
     def grain(self, grain='', incremental=False, inplace=False):
-        ''' TODO - Make incremental work '''
+        """Changes the grain of a cumulative triangle.
+
+        Parameters
+        ----------
+        grain : str
+            The grain to which you want your triangle converted, specified as
+            'O<x>D<y>' where <x> and <y> can take on values of `['Y', 'Q', 'M']`
+            For example, 'OYDY' for Origin Year/Development Year, 'OQDM' for
+            Origin quarter, etc.
+        incremental : bool
+            Not implemented yet
+        inplace : bool
+            Whether to mutate the existing Triangle instance or return a new one.
+
+        Returns
+        -------
+            Triangle
+        """
         if inplace:
             origin_grain = grain[1:2]
             development_grain = grain[-1]

chainladder/methods/capecod.py

@@ -10,34 +10,37 @@
 
 
 class CapeCod(MethodBase):
+    """Applies the CapeCod technique to triangle **X**
+
+    Parameters
+    ----------
+    X : Triangle
+        The data used to compute the mean and standard deviation
+        used for later scaling along the features axis.
+    y : None
+        Ignored
+    sample_weight : Triangle
+        Required exposure to be used in the calculation.
+
+    Attributes
+    ----------
+    triangle :
+        returns **X**
+    ultimate_ :
+        The ultimate losses per the method
+    ibnr_ :
+        The IBNR per the method
+    apriori_ :
+        The trended apriori vector developed by the Cape Cod Method
+    detrended_apriori_ :
+        The detrended apriori vector developed by the Cape Cod Method
+    """
+
     def __init__(self, trend=0, decay=1):
         self.trend = trend
         self.decay = decay
 
     def fit(self, X, y=None, sample_weight=None):
-        """Applies the CapeCod technique to triangle **X**
-
-        Parameters
-        ----------
-        X : Triangle
-            The data used to compute the mean and standard deviation
-            used for later scaling along the features axis.
-        y : None
-            Ignored
-        sample_weight : Triangle
-            Required exposure to be used in the calculation.
-
-        Attributes
-        ----------
-        triangle :
-            returns **X**
-        ultimate_ :
-            The ultimate losses per the method
-        ibnr_ :
-            The IBNR per the method
-        apriori_ :
-            The apriori vector developed by the Cape Cod Method
-        """
         super().fit(X, y, sample_weight)
         self.sample_weight_ = sample_weight
         latest = self.X_.latest_diagonal.triangle
@@ -49,19 +52,21 @@ def fit(self, X, y=None, sample_weight=None):
         trend_exponent = exposure.shape[-2]-np.arange(exposure.shape[-2])-1
         trend_array = (1+self.trend)**(trend_exponent)
         trend_array = self.X_.expand_dims(np.expand_dims(trend_array, -1))
-        decay_matrix = self.decay * \
+        decay_matrix = self.decay ** \
             (np.abs(np.expand_dims(np.arange(exposure.shape[-2]), 0).T -
              np.expand_dims(np.arange(exposure.shape[-2]), 0)))
         decay_matrix = self.X_.expand_dims(decay_matrix)
-        weighted_exposure = reported_exposure * decay_matrix
-        trended_ultimate = (latest*trend_array)/(reported_exposure)
+        weighted_exposure = np.swapaxes(reported_exposure, -1, -2)*decay_matrix
+        trended_ultimate = np.swapaxes((latest*trend_array) /
+                                       reported_exposure, -1, -2)
         apriori = np.sum(weighted_exposure*trended_ultimate, -1) / \
             np.sum(weighted_exposure, -1)
-
         obj.triangle = np.expand_dims(apriori, -1)
         obj.ddims = ['Apriori']
         self.apriori_ = copy.deepcopy(obj)
         detrended_ultimate = self.apriori_.triangle/trend_array
+        self.detrended_apriori_ = copy.deepcopy(obj)
+        self.detrended_apriori_.tiangle = detrended_ultimate
         ibnr = detrended_ultimate*(1-1/cdf)*exposure
         obj.triangle = latest + ibnr
         obj.ddims = ['Ultimate']

chainladder/methods/tests/test_capecod.py

@@ -0,0 +1,8 @@
+import chainladder as cl
+
+def test_struhuss():
+    X = cl.load_dataset('cc_sample')['loss']
+    X = cl.TailConstant(tail=1/0.85).fit_transform(cl.Development().fit_transform(X))
+    sample_weight = cl.load_dataset('cc_sample')['exposure'].latest_diagonal
+    ibnr = int(cl.CapeCod(trend=0.07, decay=0.75).fit(X, sample_weight=sample_weight).ibnr_.sum()[0])
+    assert ibnr == 17052

chainladder/utils/data/cc_sample.csv

@@ -0,0 +1,16 @@
+,origin,development,loss,exposure
+0,1993,1993,1058.823529,7000
+1,1994,1994,1333.333333,8000
+2,1995,1995,2000.0,9000
+3,1996,1996,2000.0,10000
+4,1997,1997,2800.0,11000
+5,1993,1994,1905.8823530000002,7000
+6,1994,1995,2400.0,8000
+7,1995,1996,3600.0,9000
+8,1996,1997,3600.0,10000
+9,1993,1995,2541.176471,7000
+10,1994,1996,3200.0,8000
+11,1995,1997,4800.0,9000
+12,1993,1996,3176.4705879999997,7000
+13,1994,1997,4000.0,8000
+14,1993,1997,3600.0,7000

chainladder/utils/utility_functions.py

@@ -32,9 +32,11 @@ def load_dataset(key):
         development = 'DevelopmentYear'
         index = ['GRNAME', 'LOB']
         columns = ['IncurLoss', 'CumPaidLoss', 'BulkLoss', 'EarnedPremDIR',
-                  'EarnedPremCeded', 'EarnedPremNet']
+                   'EarnedPremCeded', 'EarnedPremNet']
     if key.lower() in ['liab', 'auto']:
         index = ['lob']
+    if key.lower() in ['cc_sample']:
+        columns = ['loss', 'exposure']
     df = pd.read_csv(os.path.join(path, 'data', key.lower() + '.csv'))
     return Triangle(df, origin=origin, development=development,
                     columns=columns, index=index)

chainladder/workflow/gridsearch.py

@@ -51,7 +51,7 @@ def __init__(self, estimator, param_grid, scoring, verbose=0,
         self.verbose = verbose
         self.error_score = error_score
 
-    def fit(self, X):
+    def fit(self, X, y=None, sample_weight=None):
         if type(self.scoring) is not dict:
             scoring = dict(score=self.scoring)
         else:
@@ -60,7 +60,7 @@ def fit(self, X):
         results_ = []
         for num, item in enumerate(grid):
             est = copy.deepcopy(self.estimator).set_params(**item)
-            model = est.fit(X)
+            model = est.fit(X, y, sample_weight)
             for score in scoring.keys():
                 item[score] = scoring[score](model)
             results_.append(item)

docs/auto_examples/index.rst

@@ -107,6 +107,26 @@ Gallery of Chainladder Functionality
 
    /auto_examples/plot_bf_apriori_from_cl
 
+.. raw:: html
+
+    <div class="sphx-glr-thumbcontainer" tooltip="This example demonstrates the relationship between the Chainladder and BornhuetterFerguson meth...">
+
+.. only:: html
+
+    .. figure:: /auto_examples/images/thumb/sphx_glr_plot_benktander_thumb.png
+
+        :ref:`sphx_glr_auto_examples_plot_benktander.py`
+
+.. raw:: html
+
+    </div>
+
+
+.. toctree::
+   :hidden:
+
+   /auto_examples/plot_benktander
+
 .. raw:: html
 
     <div class="sphx-glr-thumbcontainer" tooltip="This example demonstrates the typical way you&#x27;d ingest data into a Triangle. Data in tabular fo...">
@@ -146,6 +166,26 @@ Gallery of Chainladder Functionality
    :hidden:
 
    /auto_examples/plot_munich
+
+.. raw:: html
+
+    <div class="sphx-glr-thumbcontainer" tooltip="This example demonstrates the usage of the deterministic CapeCod method and shows the sensitivi...">
+
+.. only:: html
+
+    .. figure:: /auto_examples/images/thumb/sphx_glr_plot_capecod_thumb.png
+
+        :ref:`sphx_glr_auto_examples_plot_capecod.py`
+
+.. raw:: html
+
+    </div>
+
+
+.. toctree::
+   :hidden:
+
+   /auto_examples/plot_capecod
 .. raw:: html
 
     <div style='clear:both'></div>

docs/auto_examples/plot_benktander.ipynb

@@ -0,0 +1,54 @@
+{
+  "cells": [
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {
+        "collapsed": false
+      },
+      "outputs": [],
+      "source": [
+        "%matplotlib inline"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "\n====================================================================\nBenktander: Relationship between Chainladder and BornhuetterFerguson\n====================================================================\n\nThis example demonstrates the relationship between the Chainladder and\nBornhuetterFerguson methods by way fo the Benktander model. Each is a\nspecial case of the Benktander model where `n_iters = 1` for BornhuetterFerguson\nand as `n_iters` approaches infinity yields the chainladder.  As `n_iters`\nincreases the apriori selection becomes less relevant regardless of initial\nchoice.\n\n"
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {
+        "collapsed": false
+      },
+      "outputs": [],
+      "source": [
+        "import chainladder as cl\n\nimport seaborn as sns\nimport matplotlib.pyplot as plt\nsns.set_style('whitegrid')\n\n# Load Data\nclrd = cl.load_dataset('clrd')\nmedmal_paid = clrd.groupby('LOB').sum().loc['medmal']['CumPaidLoss']\nmedmal_prem = clrd.groupby('LOB').sum().loc['medmal']['EarnedPremDIR'].latest_diagonal\nmedmal_prem.rename(development='premium')\n\n# Generate LDFs and Tail Factor\nmedmal_paid = cl.Development().fit_transform(medmal_paid)\nmedmal_paid = cl.TailCurve().fit_transform(medmal_paid)\n\n# Benktander Model\nbenk = cl.Benktander()\n\n# Prep Benktander Grid Search with various assumptions, and a scoring function\nparam_grid = dict(n_iters=list(range(1,100,2)),\n                  apriori=[0.50, 0.75, 1.00])\nscoring = {'IBNR':lambda x: x.ibnr_.sum()[0]}\ngrid = cl.GridSearch(benk, param_grid, scoring=scoring)\n# Perform Grid Search\ngrid.fit(medmal_paid, sample_weight=medmal_prem)\n\n# Plot data\ngrid.results_.pivot(index='n_iters', columns='apriori', values='IBNR').plot()\nplt.title('Benktander convergence to Chainladder')\ng = plt.ylabel('IBNR')"
+      ]
+    }
+  ],
+  "metadata": {
+    "kernelspec": {
+      "display_name": "Python 3",
+      "language": "python",
+      "name": "python3"
+    },
+    "language_info": {
+      "codemirror_mode": {
+        "name": "ipython",
+        "version": 3
+      },
+      "file_extension": ".py",
+      "mimetype": "text/x-python",
+      "name": "python",
+      "nbconvert_exporter": "python",
+      "pygments_lexer": "ipython3",
+      "version": "3.6.3"
+    }
+  },
+  "nbformat": 4,
+  "nbformat_minor": 0
+}

docs/auto_examples/plot_benktander.py

@@ -0,0 +1,43 @@
+"""
+====================================================================
+Benktander: Relationship between Chainladder and BornhuetterFerguson
+====================================================================
+
+This example demonstrates the relationship between the Chainladder and
+BornhuetterFerguson methods by way fo the Benktander model. Each is a
+special case of the Benktander model where `n_iters = 1` for BornhuetterFerguson
+and as `n_iters` approaches infinity yields the chainladder.  As `n_iters`
+increases the apriori selection becomes less relevant regardless of initial
+choice.
+"""
+import chainladder as cl
+
+import seaborn as sns
+import matplotlib.pyplot as plt
+sns.set_style('whitegrid')
+
+# Load Data
+clrd = cl.load_dataset('clrd')
+medmal_paid = clrd.groupby('LOB').sum().loc['medmal']['CumPaidLoss']
+medmal_prem = clrd.groupby('LOB').sum().loc['medmal']['EarnedPremDIR'].latest_diagonal
+medmal_prem.rename(development='premium')
+
+# Generate LDFs and Tail Factor
+medmal_paid = cl.Development().fit_transform(medmal_paid)
+medmal_paid = cl.TailCurve().fit_transform(medmal_paid)
+
+# Benktander Model
+benk = cl.Benktander()
+
+# Prep Benktander Grid Search with various assumptions, and a scoring function
+param_grid = dict(n_iters=list(range(1,100,2)),
+                  apriori=[0.50, 0.75, 1.00])
+scoring = {'IBNR':lambda x: x.ibnr_.sum()[0]}
+grid = cl.GridSearch(benk, param_grid, scoring=scoring)
+# Perform Grid Search
+grid.fit(medmal_paid, sample_weight=medmal_prem)
+
+# Plot data
+grid.results_.pivot(index='n_iters', columns='apriori', values='IBNR').plot()
+plt.title('Benktander convergence to Chainladder')
+g = plt.ylabel('IBNR')

docs/auto_examples/plot_benktander.py.md5

@@ -0,0 +1 @@
+7d348baf6a5767e0aa7b52f411601199