various other fixes

convoys/__init__.py

@@ -28,12 +28,18 @@ def get_arrays(groups, data, t_converter):
     G, B, T = [], [], []
     group2j = dict((group, j) for j, group in enumerate(groups))
     for group, created_at, converted_at, now in data:
+        if created_at is None:
+            print('created at is None')
+            continue
         if converted_at is not None and converted_at <= created_at:
             print('created at', created_at, 'but converted at', converted_at)
             continue
         if now < created_at:
             print('created at', created_at, 'but now is', now)
             continue
+        if converted_at is not None and now < converted_at:
+            print('converted at', converted_at, 'but now is', now)
+            continue
         if group in group2j:
             G.append(group2j[group])
             B.append(converted_at is not None)

convoys/multi.py

@@ -23,8 +23,11 @@ def _get_x(self, group):
         x[group] = 1
         return x
 
-    def cdf(self, group, t, *args, **kwargs):
-        return self.base_model.cdf(self._get_x(group), t, *args, **kwargs)
+    def cdf(self, group, *args, **kwargs):
+        return self.base_model.cdf(self._get_x(group), *args, **kwargs)
+
+    def rvs(self, group, *args, **kwargs):
+        return self.base_model.rvs(self._get_x(group), *args, **kwargs)
 
 
 class SingleToMulti(MultiModel):

convoys/regression.py

@@ -115,6 +115,8 @@ def rvs(self, x, n_curves=1, n_samples=1, T=None):
         # T is optional and means we already observed non-conversion until T
         if T is None:
             T = numpy.zeros((n_curves, n_samples))
+        else:
+            assert T.shape == (n_curves, n_samples)
         a = LinearCombination.sample(self.params['a'], x, 1, n_curves)
         b = LinearCombination.sample(self.params['b'], x, 1, n_curves)
         B = numpy.zeros((n_curves, n_samples), dtype=numpy.bool)
@@ -123,15 +125,15 @@ def rvs(self, x, n_curves=1, n_samples=1, T=None):
             z = numpy.random.uniform(size=(n_samples,))
             cdf_now = expit(b) * gammainc(
                 self.params['k'],
-                (T[i]*numpy.exp(a))**self.params['p'])
+                numpy.multiply.outer(T[i], numpy.exp(a))**self.params['p'])
             cdf_final = expit(b)
             adjusted_z = cdf_now + (1 - cdf_now) * z
             B[i] = (adjusted_z < cdf_final)
             y = adjusted_z / cdf_final
             x = gammaincinv(self.params['k'], y)
             # x = (t * exp(a))**p
             C[i] = x**(1./self.params['p']) / numpy.exp(a)
-            C[~B] = 0
+            C[i][~B[i]] = 0
 
         return B, C