def test_no_kids_over_89(data_answers_name):
cons_name = 'no_kids_over_89'
# Get the constraint, data with answers and schema
c, data_answers, schema = ctu.getConstraintDataAnswersSchema(schemaname, data_answers_name, ctu.getAllImplementedInvariantNames(schemaname), cons_name)
assert c.sign == "="
assert c.rhs[0] == 0
assert ctu.lhsFromMatrix(c, data_answers["hist"][0])[0] == data_answers[cons_name]
# Selecting relatives
# 2 'Biological Son/Daughter',
# 3 'Adopted Son/Daughter',
# 4 'Stepson/Stepdaughter',
# 9 'Son/Daughter-in-law',
ctu.assertKronFactor(c, CC.ATTR_REL, np.array([[False, False, True, True, True] + [False] * 4 + [True] + [False] * 33]), schema)
# # This one also includes 7 for "Grandchild"
# ctu.assertKronFactor(c, CC.ATTR_HHGQ, np.array([[False, False, True, True, True] + [False, False, True, False, True] + [False] * 33]), schema)
# Selecting people over 89 (First 90 False, the rest is True)
ctu.assertKronFactor(c, CC.ATTR_AGE, np.array([[False] * 90 + [True] * 26]), schema)
# Everyone on all other axes
for dimname in ctu.excludeAx((CC.ATTR_REL, CC.ATTR_AGE), schema):
ctu.assertKronFactorSelectAll(c, dimname, schema)
def test_nurse_nva_0(data_answers_name):
cons_name = 'nurse_nva_0'
# Get the constraint, data with answers and schema
c, data_answers, schema = ctu.getConstraintDataAnswersSchema(
schemaname, data_answers_name,
ctu.getAllImplementedInvariantNames(schemaname), cons_name)
# Check sign and right hand side which is zero (in constraint) since it's a structural zero.
assert c.sign == "="
assert c.rhs[0] == 0
# If constraint query matrix is applied to the data it doesn't have to be zero
assert ctu.lhsFromMatrix(
c, data_answers["hist"][0])[0] == data_answers[cons_name]
# Check Kronecker factors
# Selecting only nursing homes (index=3)
ctu.assertKronFactorSingle(c, CC.ATTR_HHGQ, 3, schema)
# Selecting only non-voting age (index=0)
ctu.assertKronFactorSingle(c, CC.ATTR_VOTING, 0, schema)
# Selecting everyone on other axes
for dimname in ctu.excludeAx((CC.ATTR_HHGQ, CC.ATTR_VOTING), schema):
ctu.assertKronFactorSelectAll(c, dimname, schema)
def test_hhgq_va(data_answers_name, cons_name):
## Get the constraint, data with answers and schema
c, data_answers, schema = ctu.getConstraintDataAnswersSchema(
schemaname, data_answers_name,
ctu.getAllImplementedInvariantNames(schemaname), cons_name)
# Check sign and right hand side
assert c.sign == "le" if cons_name == 'hhgq_va_ub' else "ge"
assert np.array_equal(c.rhs, data_answers[cons_name])
if c.sign == "le":
assert np.all(ctu.lhsFromMatrix(c, data_answers["hist"][0]) <= c.rhs)
else:
assert np.all(ctu.lhsFromMatrix(c, data_answers["hist"][0]) >= c.rhs)
# Check Kronecker factors
# Each column of the HHGQ matrix selects that GQ type (so only one True value -- on the diagonal)
ctu.assertKronFactorDiagonal(c, CC.ATTR_HHGQ, schema)
# For non-voting, select first AGECAT_VA_CUT categories, for voting select the other categories
ctu.assertKronFactor(
c, CC.ATTR_AGECAT,
np.array([[True] * AGECAT_VA_CUT + [False] *
(ctu.dimSch(CC.ATTR_AGECAT, schema) - AGECAT_VA_CUT),
[False] * AGECAT_VA_CUT + [True] *
(ctu.dimSch(CC.ATTR_AGECAT, schema) - AGECAT_VA_CUT)]),
schema)
# Selecting everyone on all other axes
for dimname in (CC.ATTR_SEX, CC.ATTR_HISP, CC.ATTR_CENRACE):
ctu.assertKronFactorSelectAll(c, dimname, schema)
def test_householder_ub(data_answers_name):
cons_name = "householder_ub"
inv_names = list(ctu.getAllImplementedInvariantNames(schemaname))
c, data_answers, schema = ctu.getConstraintDataAnswersSchema(schemaname, data_answers_name, inv_names, cons_name)
# Check right hand side
assert np.all(c.rhs == data_answers[cons_name])
# Check left hand side
assert np.all(ctu.lhsFromMatrix(c, data_answers["hist"][0]) == data_answers[cons_name + "_lhs"])
# Check sign
assert c.sign == "le"
assert np.all(ctu.lhsFromMatrix(c, data_answers["hist"][0]) <= c.rhs)
dim_hhgq = ctu.dimSch(CC.ATTR_RELGQ, schema)
kf_hhgq = c.query.queries[0].kronFactors()[ctu.indSch(CC.ATTR_RELGQ, schema)]
assert kf_hhgq.shape == (1, dim_hhgq)
# True for 0,1 -- householder living alone and householder not living alone
assert np.array_equal(kf_hhgq[0, :], np.array([True] * 2 + [False] * (dim_hhgq - 2)))
# Selecting everyone on all other axes
for dimname in ctu.excludeAx((CC.ATTR_RELGQ,), schema):
ctu.assertKronFactorSelectAll(c, dimname, schema)
def test_no_parents_under_30(data_answers_name):
cons_name = 'no_parents_under_30'
# Get the constraint, data with answers and schema
c, data_answers, schema = ctu.getConstraintDataAnswersSchema(schemaname, data_answers_name, ctu.getAllImplementedInvariantNames(schemaname), cons_name)
assert c.sign == "="
assert c.rhs[0] == 0
assert ctu.lhsFromMatrix(c, data_answers["hist"][0])[0] == data_answers[cons_name]
# Selecting relatives "6" (parents) and "8" (parents-in-law)
ctu.assertKronFactor(c, CC.ATTR_REL, np.array([[False] * 6 + [True, False, True] + [False] * 34]), schema)
# Selecting people under 30 (First 30 True, the rest is False)
ctu.assertKronFactor(c, CC.ATTR_AGE, np.array([[True] * 30 + [False] * 86]), schema)
# Everyone on all other axes
for dimname in ctu.excludeAx((CC.ATTR_REL, CC.ATTR_AGE), schema):
ctu.assertKronFactorSelectAll(c, dimname, schema)
def test_no_vacant(data_answers_name, schemaname):
cons_name = 'no_vacant'
# Get the constraint, data with answers and schema
c, data_answers, schema = ctu.getConstraintDataAnswersSchema(
schemaname, data_answers_name,
ctu.getAllImplementedInvariantNames(schemaname), cons_name)
assert c.sign == "="
assert c.rhs[0] == 0
assert ctu.lhsFromMatrix(
c, data_answers["hist"][0])[0] == data_answers[cons_name]
# Select only size 0
ctu.assertKronFactor(c, SIZE, np.array([[True] + [False] * 7]), schema)
# Everyone on all other axes
for dimname in ctu.excludeAx((SIZE, ), schema):
ctu.assertKronFactorSelectAll(c, dimname, schema)
def test_hhgq_total(data_answers_name, tot_inv, cons_name):
# Create invariant names, remove 'tot' if checking without total
inv_names = list(ctu.getAllImplementedInvariantNames(schemaname))
if not tot_inv:
inv_names.remove('tot')
# Get the constraint, data with answers and schema
c, data_answers, schema = ctu.getConstraintDataAnswersSchema(schemaname, data_answers_name, inv_names, cons_name)
# Check sign, <= for upper bound, >= for lower
assert c.sign == "le" if cons_name == 'hhgq_total_ub' else "ge"
# Check right hand side
if tot_inv:
assert np.array_equal(c.rhs, data_answers[cons_name])
else:
assert np.array_equal(c.rhs, data_answers[cons_name + "_no_tot"])
# Check right hand side from matrix application to the data
if c.sign == "le":
assert np.all(ctu.lhsFromMatrix(c, data_answers["hist"][0]) <= c.rhs)
else:
assert np.all(ctu.lhsFromMatrix(c, data_answers["hist"][0]) >= c.rhs)
# Should get 1 + (schDim(HHGQ) - NUM_HH_CATS) rows x schDim(HHGQ) columns
dim_hhgq = ctu.dimSch(CC.ATTR_REL, schema)
dim_gq = dim_hhgq - NUM_HH_CATS
kf_hhgq = c.query.queries[0].kronFactors()[ctu.indSch(CC.ATTR_REL, schema)]
assert kf_hhgq.shape == (dim_gq + 1, dim_hhgq)
# First row is selecting first NUM_HH_CATS, and dropping all the GQs, so first NUM_HH_CATS values True, then False
assert np.array_equal(kf_hhgq[0, :], np.array([True] * NUM_HH_CATS + [False] * dim_gq))
# The rest rows drop first NUM_HH_CATS values
assert np.array_equal(kf_hhgq[1:, :NUM_HH_CATS], np.broadcast_to(np.array([False]), (dim_gq, NUM_HH_CATS)))
# and select only the corresponding GQ type, leaving a diagonal of True in lower right corner of the matrix
assert np.array_equal(kf_hhgq[1:, NUM_HH_CATS:], np.diag([True]*dim_gq))
# Selecting everyone on all other axes
for dimname in ctu.excludeAx((CC.ATTR_REL, ), schema):
ctu.assertKronFactorSelectAll(c, dimname, schema)
def test_parents_ub(cons_name_rel_index, data_answers_name):
cons_name, rel_index = cons_name_rel_index
inv_names = list(ctu.getAllImplementedInvariantNames(schemaname))
c, data_answers, schema = ctu.getConstraintDataAnswersSchema(schemaname, data_answers_name, inv_names, cons_name)
# Check right hand side
assert np.all(c.rhs == data_answers[cons_name])
# Check left hand side
assert np.all(ctu.lhsFromMatrix(c, data_answers["hist"][0]) == data_answers[cons_name + "_lhs"])
# Check sign
assert c.sign == "le"
assert np.all(ctu.lhsFromMatrix(c, data_answers["hist"][0]) <= c.rhs)
# Parents are 10, parent-in-law are 12
ctu.assertKronFactor(c, CC.ATTR_RELGQ, np.array([[False] * rel_index + [True] + [False] * (42-rel_index-1)]), schema)
# Selecting everyone on all other axes
for dimname in ctu.excludeAx((CC.ATTR_RELGQ, ), schema):
ctu.assertKronFactorSelectAll(c, dimname, schema)
def test_no_refhh_under_15(data_answers_name):
cons_name = 'no_refhh_under_15'
# Get the constraint, data with answers and schema
c, data_answers, schema = ctu.getConstraintDataAnswersSchema(schemaname, data_answers_name, ctu.getAllImplementedInvariantNames(schemaname), cons_name)
assert c.sign == "="
assert c.rhs[0] == 0
assert ctu.lhsFromMatrix(c, data_answers["hist"][0])[0] == data_answers[cons_name]
# Selecting relatives
# 'Householder (0)', 'Husband/Wife (1)', 'Father/Mother (6)', 'Parent-in-law (8)', 'Son/Daughter-in-law (9)', 'Housemate, Roommate (12)',
# 'Unmarried Partner (13)'
age_cut = 15
ans = np.array([[False] * ctu.dimSch(CC.ATTR_REL, schema)])
ans[0, [0, 1, 6, 8, 9, 12, 13]] = True
ctu.assertKronFactor(c, CC.ATTR_REL, ans, schema)
# Selecting people under age_cut (First age_cut True, the rest are False)
ctu.assertKronFactor(c, CC.ATTR_AGE, np.array([[True] * age_cut + [False] * (ctu.dimSch(CC.ATTR_AGE, schema) - age_cut)]), schema)
# Everyone on all other axes
for dimname in ctu.excludeAx((CC.ATTR_REL, CC.ATTR_AGE), schema):
ctu.assertKronFactorSelectAll(c, dimname, schema)
def test_voting_age(data_answers_name):
cons_name = 'voting_age'
## Get the constraint, data with answers and schema
c, data_answers, schema = ctu.getConstraintDataAnswersSchema(
schemaname, data_answers_name,
ctu.getAllImplementedInvariantNames(schemaname), cons_name)
# Check sign and right hand side, in constraint and from query matrix applied to the data
assert c.sign == "="
assert c.rhs[0] == data_answers[cons_name]
assert ctu.lhsFromMatrix(
c, data_answers["hist"][0])[0] == data_answers[cons_name]
# Check Kronecker factors
# The values below AGECAT_VA_CUT are non-voting, selecting the other ones
ctu.assertKronFactorMiltiple(
c, CC.ATTR_AGECAT,
range(AGECAT_VA_CUT, ctu.dimSch(CC.ATTR_AGECAT, schema)), schema)
# Selecting everyone on other axes
for dimname in (CC.ATTR_HHGQ, CC.ATTR_SEX, CC.ATTR_HISP, CC.ATTR_CENRACE):
ctu.assertKronFactorSelectAll(c, dimname, schema)
def test_over100(data_answers_name):
cons_name = "people100Plus_ub"
inv_names = list(ctu.getAllImplementedInvariantNames(schemaname))
c, data_answers, schema = ctu.getConstraintDataAnswersSchema(schemaname, data_answers_name, inv_names, cons_name)
# Check right hand side
assert np.all(c.rhs == data_answers[cons_name])
# Check left hand side
assert np.all(ctu.lhsFromMatrix(c, data_answers["hist"][0]) == data_answers[cons_name + "_lhs"])
# Check sign
assert c.sign == "le"
assert np.all(ctu.lhsFromMatrix(c, data_answers["hist"][0]) <= c.rhs)
# People in households (REL part, up to 18)
ctu.assertKronFactor(c, CC.ATTR_RELGQ, np.array([[True] * 18 + [False] * 24]), schema)
# People over 100 years old
ctu.assertKronFactor(c, CC.ATTR_AGE, np.array([[False] * 100 + [True] * 16]), schema)
# Selecting everyone on all other axes
for dimname in ctu.excludeAx((CC.ATTR_RELGQ, CC.ATTR_AGE), schema):
ctu.assertKronFactorSelectAll(c, dimname, schema)
def test_living_alone(data_answers_name, schemaname):
cons_name = 'living_alone'
# Get the constraint, data with answers and schema
cs, data_answers, schema = ctu.getConstraintsDataAnswersSchema(
schemaname, data_answers_name,
ctu.getAllImplementedInvariantNames(schemaname), (cons_name, ))
# Check size 1, and all rhs being zero
for c in cs.values():
assert c.sign == "="
assert c.rhs[0] == 0
# Select only size 1
ctu.assertKronFactor(c, SIZE,
np.array([[False] + [True] + [False] * 6]),
schema)
name = "living_alone_gt1"
assert ctu.lhsFromMatrix(cs[name],
data_answers["hist"][0])[0] == data_answers[name]
# Select hhtypes implying > 1 person
ans = [True] * 24
ans[18] = False
ctu.assertKronFactor(cs[name], HHTYPE, np.array([ans]), schema)
# Everyone on all other axes
for dimname in ctu.excludeAx((SIZE, HHTYPE), schema):
ctu.assertKronFactorSelectAll(cs[name], dimname, schema)
name = "living_alone_multi"
assert ctu.lhsFromMatrix(cs[name],
data_answers["hist"][0])[0] == data_answers[name]
# Select multi=1
ctu.assertKronFactor(cs[name], MULTI, np.array([[False, True]]), schema)
# Everyone on all other axes
for dimname in ctu.excludeAx((SIZE, MULTI), schema):
ctu.assertKronFactorSelectAll(cs[name], dimname, schema)
for eld in [0, 1, 2, 3]:
name = f"living_alone_eld{eld}"
assert ctu.lhsFromMatrix(
cs[name], data_answers["hist"][0])[0] == data_answers[name]
# Select multi=0
ctu.assertKronFactor(cs[name], MULTI, np.array([[True, False]]),
schema)
# Select hhtypes implying 1 person
ans = [False] * 24
ans[18] = True
ctu.assertKronFactor(cs[name], HHTYPE, np.array([ans]), schema)
# select elderly not "eld"
ctu.assertKronFactor(
cs[name], ELDERLY,
np.array([[True] * eld + [False] + [True] * (3 - eld)]), schema)
# select age under 60, 60-65, 65-75 or over 75
ctu.assertKronFactor(
cs[name], HHAGE,
np.array([[not (bool(eld))] * 5 +
list(np.array([1, 2, 3]) == eld) + [eld == 3]]), schema)
# Everyone on all other axes
for dimname in ctu.excludeAx((SIZE, HHTYPE, MULTI, HHAGE, ELDERLY),
schema):
ctu.assertKronFactorSelectAll(cs[name], dimname, schema)