def test_size_distribution(self):
"""
This test may fail 1 out of 10,000 times because it tests a sample
mean against a population mean
"""
pop_avg = 100
pop_std_dev = 10
z_for_99_point_99_confidence = 3.8906
n = 20
classrooms = [
graph_generation.new_classroom(
# add 1 for the teacher
avg_num_students=pop_avg + 1,
variation_coeff=pop_std_dev * 1.0 / pop_avg
)
for c in range(n)
]
mean_users = mean([len(c) for c in classrooms])
# We generate the 99.99% confidence interval for this sample mean
max_distance = z_for_99_point_99_confidence * pop_std_dev / sqrt(n)
self.assertTrue(
abs(mean_users - pop_avg) < max_distance,
msg='classroom size exceeded expected bounds (may happen 1 out of 10,000 times)'
)
def reset_graph():
"""Resets THE graph based on parameters"""
global graph
num_schools = dict_int(request.form, 'num_schools', 5)
num_classes = dict_int(request.form, 'num_classes', 10)
class_gen = lambda: new_classroom(10, variation_coeff=1)
schools = [new_school(num_classes, 0.5, class_gen, extra_coaching_rate=0.1) for i in range(num_schools)]
# flatten schools list
graph = [user for school in schools for user in school]
return users_to_graph_json(graph)
def test_list_is_flat(self):
classroom_gen = lambda: graph_generation.new_classroom(
avg_num_students=10,
variation_coeff=0.1
)
school = graph_generation.new_school(
avg_num_classes=3,
variation_coeff=0.3,
class_gen=classroom_gen
)
for u in school:
self.assertIsInstance(u, models.User, msg='School list not flattened')
def test_teacher_exists(self):
n = 10
classrooms = [
graph_generation.new_classroom(avg_num_students=25, variation_coeff=0.1)
for c in range(n)
]
for c in classrooms:
# Make sure there's at least one user with n-1 learners
self.assertGreaterEqual(
max(len(u.learner_set) for u in c),
len(c) - 1,
msg="no teacher found"
)
