def create_report_file(Resistances, initial_resistances, drug_start_time,
param_obj, report_name, inset_days, debug):
with open(report_name, "w") as outfile:
starting_pop = inset_days[0][
dts.InsetChart.Channels.KEY_StatisticalPopulation]
# success = sft.test_binomial_95ci( initial_resistances, starting_pop, param_obj["TB_Drug_Resistance_Rate_HIV"], outfile, "???" )
success = True
progression = []
bad_msgs = []
for x in range(len(inset_days)):
inset_day = inset_days[x]
inset_mdr_prevalence = inset_day[
dts.InsetChart.Channels.KEY_MdrTbPrevalence]
stdout_resistants = Resistances[x]
if x >= drug_start_time:
progression.append(stdout_resistants)
if debug:
outfile.write("Day: {0}\n".format(x))
outfile.write(str(inset_day) + "\n")
outfile.write(
"StdOut resistants: {0}\n".format(stdout_resistants))
stdout_predicted_prevalence = stdout_resistants / float(
inset_day[dts.InsetChart.Channels.KEY_StatisticalPopulation])
if abs(inset_mdr_prevalence - stdout_predicted_prevalence) > 0.03:
bad_msgs.append(
"BAD: at timestep {0}, expected MDR prevalence: {1}, InsetChart had: {2}\n"
.format(x, stdout_predicted_prevalence,
inset_mdr_prevalence))
tb_drug_resistance_rate_hiv = param_obj["TB_Drug_Resistance_Rate_HIV"]
new_resistances = []
pre_resistance = 0
failed_count = 0
total_test = 0
for x in range(drug_start_time + 1, len(Resistances)):
resistance = Resistances[x]
new_resistance = resistance - pre_resistance
pre_resistance = resistance
new_resistances.append(new_resistance)
expected_mean = (starting_pop -
resistance) * tb_drug_resistance_rate_hiv
total_test += 1
if expected_mean >= 5: # advoid failing with too small mean
result = sft.test_binomial_99ci(
new_resistance,
starting_pop - resistance,
tb_drug_resistance_rate_hiv,
outfile,
category="time step {}".format(x + 1))
if not result:
failed_count += 1
outfile.write(
"Warning: New Resistance test fails for rate = {0} at time step {1}.\n"
.format(tb_drug_resistance_rate_hiv, x + 1))
else:
error_tolerance = 3 * math.sqrt(
tb_drug_resistance_rate_hiv *
(1 - tb_drug_resistance_rate_hiv) *
(starting_pop - resistance)) # 3 sigma
result = math.fabs(new_resistance -
expected_mean) <= error_tolerance
if not result:
failed_count += 1
outfile.write(
"Warning: New Resistance test fails for rate = {0} at time step {1}, "
"new resistance = {2}, expected mean = {3}, error tolerance = {4}.\n"
.format(tb_drug_resistance_rate_hiv, x + 1,
new_resistance, expected_mean,
error_tolerance))
if failed_count > math.ceil(total_test * 0.01):
success = False
outfile.write(
"BAD: test failed {0} times out of {1} timestep, please check the warning message.\n"
"".format(failed_count, total_test))
if debug:
sft.plot_data(new_resistances,
title="new resistance over time",
category="new_resistance",
show=True)
series = sft.create_geometric_dis(
param_obj["TB_Drug_Resistance_Rate_HIV"],
starting_pop,
len(progression),
test_decay=False)
sft.plot_data(progression,
series,
label1="progression",
label2="geomatric dis",
xlabel="days",
ylabel="resistance",
title="progression vs geomatric",
category="progression_vs_geomatric",
show=True,
line=True)
sft.plot_cdf(progression,
series,
label1="progression",
label2="geomatric dis",
title="progression vs geomatric cdf",
category="progression_vs_geomatric_cdf",
show=True)
# success = sft.test_geometric_decay(progression, param_obj["TB_Drug_Resistance_Rate_HIV"], starting_pop, test_decay=False, report_file=outfile, debug=debug)
if len(bad_msgs) > 0:
success = False
outfile.writelines(bad_msgs)
outfile.write(sft.format_success_msg(success))
def create_report_file(param_obj, multipliers, infectiousness, report_name, debug):
with open(report_name, "w") as outfile:
success = True
if not multipliers:
outfile.write(sft.sft_no_test_data)
sigma = param_obj[Param_keys.LOGNORMAL_SCALE]
base_infectivity = param_obj[Param_keys.BASE_INFECTIVITY]
if sigma > 0:
mu = - sigma**2 / 2.0
# test log_normal distribution
success = sft.test_lognorm(multipliers,mu=mu, sigma=sigma,report_file=outfile,round = False)
# test mean_l = 1
mean_l = np.mean(multipliers)
mean_infectiousness = np.mean(infectiousness)
outfile.write("mean of the multipliers is {}, expected 1.0.\n".format(mean_l))
outfile.write("mean of the Infectiousness is {0}, while base infectivity is {1}.\n".format(mean_infectiousness,
base_infectivity))
tolerance = 2e-2
if math.fabs(mean_l - 1.0) > tolerance:
outfile.write("BAD: mean of the multipliers is {}, expected 1.0.\n".format(mean_l))
success = False
# plotting
size = len(multipliers)
outfile.write("size is {}\n".format(size))
scale = math.exp(mu)
dist_lognormal = stats.lognorm.rvs(sigma, 0, scale, size)
sft.plot_data(multipliers, dist_lognormal,
label1="Emod", label2="Scipy",
ylabel="Multiplier", xlabel="data point",
category="Emod_vs_Scipy",
title="Emod_vs_Scipy, sigma = {}".format(sigma),
show=True)
sft.plot_probability(multipliers, dist_lognormal,
precision=1, label1="Emod", label2="Scipy",
category="Probability_mass_function_Emod_vs_Scipy",
title="Emod_vs_Scipy, sigma = {}".format(sigma),
show=True)
sft.plot_cdf(multipliers,dist_lognormal,label1="Emod", label2="Scipy",
category="cdf",
title="cdf, sigma = {}".format(sigma),
show=True, line = False)
if debug:
with open("scipy_data.txt", "w") as file:
for n in sorted(dist_lognormal):
file.write(str(n) + "\n")
with open("emod_data.txt", "w") as file:
for n in sorted(multipliers):
file.write(str(n) + "\n")
else:
# sigma = 0, this feature is disabled
for multiplier in multipliers:
if multiplier != 1.0:
success = False
outfile.write("BAD: multiplier is {0} when {1} set to {2}, expected 1.0.\n".format(multiplier, Param_keys.LOGNORMAL_SCALE, sigma))
# plotting
sft.plot_data(multipliers, label1="Multiplier", label2="NA",
category="Multiplier", title="Multiplier_Sigma={}".format(sigma),
ylabel="Multiplier", xlabel="data point",
show=True)
sft.plot_data(infectiousness, label1="Infectiousness",
label2="NA",category="Infectiousness",
title="Infectiousness_Sigma={0}_BaseInfectivity={1}".format(sigma,base_infectivity),
ylabel="Infectiousness",xlabel="data point",
show=True)
outfile.write(sft.format_success_msg(success))
if debug:
print "SUMMARY: Success={0}\n".format(success)
return success
def create_report_file(drug_start_timestep, inactivation_times, active_count, inactivations, drug_inactivation_rate, report_name, debug = False):
with open(report_name, "w") as outfile:
success = True
# ks exponential test doesn't work very well with large rate, use chi squared test instead.
# while rate is small ks test for exponential distribution is more sensitive to catch the difference
if drug_inactivation_rate < 0.1:
outfile.write( "Testing inactivation times as draws from exponential distrib with rate {0}. "
"Dataset size = {1}.\n".format( drug_inactivation_rate, len( inactivation_times ) ) )
success = dtk_sft.test_exponential( inactivation_times, drug_inactivation_rate, outfile, integers=True,
roundup=True, round_nearest=False )
if not success:
outfile.write("BAD: ks test for rate {} is False.\n".format(drug_inactivation_rate))
size = len(inactivation_times)
scale = 1.0 / drug_inactivation_rate
dist_exponential_np = numpy.random.exponential(scale, size)
dist_exponential_np = [math.ceil(x) for x in dist_exponential_np]
dtk_sft.plot_data_sorted(inactivation_times, dist_exponential_np,
label1="test times", label2="numpy data",
title="inactivation_times_actual_vs_numpy",
xlabel="data points", ylabel="Inactivation times",
category="inactivation_times", show = True, line = True, overlap=True)
dtk_sft.plot_cdf(inactivation_times, dist_exponential_np,
label1="test times", label2="numpy data",
title="inactivation_times_cdf",
xlabel="days", ylabel="probability",
category="inactivation_times_cdf", show = True)
dtk_sft.plot_probability(inactivation_times, dist_exponential_np,
label1="test times", label2="numpy data",
title="inactivation_times_pdf",
xlabel="days", ylabel="probability",
category="inactivation_times_pdf", show = True)
else:
outfile.write("Testing inactivation count per day with rate {0}. \n".format( drug_inactivation_rate) )
expected_inactivation = []
for t in range( len(inactivations)):
if t < drug_start_timestep :
if inactivations[t] > 0:
success = False
outfile.write("BAD: expected no inactivations on drugs before day {0}, get {1} cases at timestep {2}.\n"
"".format(drug_start_timestep , inactivations[t], t))
elif active_count[t] > 0:
expected_inactivation.append(drug_inactivation_rate * active_count[t])
if len(inactivations) <= len(expected_inactivation) + drug_start_timestep:
test_inactivation_dates = inactivations[drug_start_timestep+1:]
expected_inactivation = expected_inactivation[:len(test_inactivation_dates)]
else:
test_inactivation_dates = inactivations[drug_start_timestep + 1:drug_start_timestep + 1 + len(expected_inactivation)]
#print (len(inactivations), len(test_inactivation_dates), len(expected_inactivation))
#print (test_inactivation_dates, expected_inactivation)
dtk_sft.plot_data(test_inactivation_dates, expected_inactivation,
label1="actual inactivation", label2="expected inactivation",
title="inactivation per day",
xlabel="date after drug start day", ylabel="inactivation per day",
category="inactivation_counts", show=True, line=True, overlap=True, sort=False)
chi_result = dtk_sft.test_multinomial(dist=test_inactivation_dates, proportions=expected_inactivation,
report_file=outfile, prob_flag=False)
if not chi_result:
success = False
outfile.write("BAD: Chi-squared test reuslt is False.\n")
outfile.write(dtk_sft.format_success_msg(success))
if debug:
print(dtk_sft.format_success_msg(success))
return success
def create_report_file(param_obj, multipliers, infectiousness, report_name, debug):
with open(report_name, "w") as outfile:
success = True
if not multipliers:
outfile.write(sft.sft_no_test_data)
sigma = param_obj[Param_keys.LOGNORMAL_SCALE]
base_infectivity = param_obj[Param_keys.BASE_INFECTIVITY]
if sigma > 0:
mu = - sigma**2 / 2.0
# test log_normal distribution
success = sft.test_lognorm(multipliers,mu=mu, sigma=sigma,report_file=outfile,round = False)
# test mean_l = 1
mean_l = np.mean(multipliers)
mean_infectiousness = np.mean(infectiousness)
outfile.write("mean of the multipliers is {}, expected 1.0.\n".format(mean_l))
outfile.write("mean of the Infectiousness is {0}, while base infectivity is {1}.\n".format(mean_infectiousness,
base_infectivity))
tolerance = 2e-2
if math.fabs(mean_l - 1.0) > tolerance:
outfile.write("BAD: mean of the multipliers is {}, expected 1.0.\n".format(mean_l))
success = False
# plotting
size = len(multipliers)
outfile.write("size is {}\n".format(size))
scale = math.exp(mu)
dist_lognormal = stats.lognorm.rvs(sigma, 0, scale, size)
sft.plot_data_sorted(multipliers, dist_lognormal,
label1="Emod", label2="Scipy",
ylabel="Multiplier", xlabel="data point",
category="Emod_vs_Scipy",
title="Emod_vs_Scipy, sigma = {}".format(sigma),
show=True)
sft.plot_probability(multipliers, dist_lognormal,
precision=1, label1="Emod", label2="Scipy",
category="Probability_mass_function_Emod_vs_Scipy",
title="Emod_vs_Scipy, sigma = {}".format(sigma),
show=True)
sft.plot_cdf(multipliers,dist_lognormal,label1="Emod", label2="Scipy",
category="cdf",
title="cdf, sigma = {}".format(sigma),
show=True, line = False)
if debug:
with open("scipy_data.txt", "w") as file:
for n in sorted(dist_lognormal):
file.write(str(n) + "\n")
with open("emod_data.txt", "w") as file:
for n in sorted(multipliers):
file.write(str(n) + "\n")
else:
# sigma = 0, this feature is disabled
for multiplier in multipliers:
if multiplier != 1.0:
success = False
outfile.write("BAD: multiplier is {0} when {1} set to {2}, expected 1.0.\n".format(multiplier, Param_keys.LOGNORMAL_SCALE, sigma))
# plotting
sft.plot_data_sorted(multipliers, label1="Multiplier", label2="NA",
category="Multiplier", title="Multiplier_Sigma={}".format(sigma),
ylabel="Multiplier", xlabel="data point",
show=True)
sft.plot_data_sorted(infectiousness, label1="Infectiousness",
label2="NA",category="Infectiousness",
title="Infectiousness_Sigma={0}_BaseInfectivity={1}".format(sigma,base_infectivity),
ylabel="Infectiousness",xlabel="data point",
show=True)
outfile.write(sft.format_success_msg(success))
if debug:
print( "SUMMARY: Success={0}\n".format(success) )
return success
def create_report_file(drug_start_timestep, disease_deaths, cum_deaths, deaths, infected_individuals, death_times, drug_mortality_rate_HIV, report_name ):
with open(report_name, "w") as outfile:
success = True
length = len(cum_deaths)
if sum(disease_deaths)==0 or sum(cum_deaths)==0 or len(death_times)==0:
success = False
outfile.write(dtk_sft.no_test_data)
for x in range(length):
if disease_deaths[x] != cum_deaths[x]:
success = False
outfile.write("BAD: at timestep {0}, disease deaths is {1} in InsetChart.json and {2} in stdout.txt.\n".format(x+1, disease_deaths[x], cum_deaths[x]))
# ks exponential test doesn't work very well with large rate, use chi squared test instead
# while rate is small ks test for exponential distribution is more sensitive to catch the difference
if drug_mortality_rate_HIV < 0.1:
outfile.write("Testing death times as draws from exponential distrib with rate {0}. "
"Dataset size = {1}.\n".format(drug_mortality_rate_HIV, len(death_times)))
ks_result = dtk_sft.test_exponential( death_times, drug_mortality_rate_HIV, report_file = outfile,
integers=True, roundup=True, round_nearest=False )
if not ks_result:
success = False
outfile.write("BAD: ks test reuslt is False.\n")
size = len(death_times)
scale = 1.0 / drug_mortality_rate_HIV
dist_exponential_np = numpy.random.exponential(scale, size)
dist_exponential_np = [math.ceil(x) for x in dist_exponential_np]
dtk_sft.plot_data_sorted(death_times, dist_exponential_np,
label1="death times", label2="numpy data",
title="death_times_actual_vs_numpy",
xlabel="data points", ylabel="death times",
category="death_times", show=True, line = True, overlap=True)
dtk_sft.plot_cdf(death_times, dist_exponential_np,
label1="death times", label2="numpy data",
title="death_times_cdf",
xlabel="days", ylabel="probability",
category="death_times_cdf", show=True)
else:
outfile.write("Testing death count per day with rate {0}. \n".format(drug_mortality_rate_HIV))
expected_mortality = []
for t in range( len(deaths)):
if t < drug_start_timestep + 1:
if deaths[t] > 0:
success = False
outfile.write("BAD: expected no disease death on drugs before day {0}, get {1} cases at timestep {2}.\n"
"".format(drug_start_timestep + 1, deaths[t], t))
elif infected_individuals[t] > 0:
expected_mortality.append(drug_mortality_rate_HIV * infected_individuals[t])
expected_mortality.pop(0) # the Infected is off by one day
test_death_dates = deaths[drug_start_timestep + 1:drug_start_timestep + 1 + len(expected_mortality)]
dtk_sft.plot_data(test_death_dates, expected_mortality,
label1="actual death", label2="expected death",
title="death per day",
xlabel="date after drug start day", ylabel="death per day",
category="death_counts", show=True, line=True, overlap=True, sort=False)
chi_result = dtk_sft.test_multinomial(dist=test_death_dates, proportions=expected_mortality,
report_file=outfile, prob_flag=False)
if not chi_result:
success = False
outfile.write("BAD: Chi-squared test reuslt is False.\n")
outfile.write(dtk_sft.format_success_msg(success))
return success