def boxplot(phenotype):
"""Draw a boxplot for the given continuous phenotype.
:param phenotype: Display a boxplot for the provided continuous phenotype.
:type phenotype: str
An exception will be raised if the required phenotype is not continuous.
"""
data, meta = _get_data_meta(phenotype)
data = data[~np.isnan(data)]
_type = types.type_str(meta["variable_type"])
if _type.subtype_of(types.Continuous):
raise REPLException("Can't draw boxplot for non-continuous variable "
"('{}').".format(phenotype))
fig, ax = plt.subplots(1, 1)
ax.boxplot(data, vert=False)
ax.set_xlabel(phenotype)
ax.set_yticklabels([])
ax.yaxis.set_ticks_position("none")
filename = "cohort_plot.png"
plt.savefig(filename, dpi=300)
return _response_from_img_filename(filename)
def info(phen_or_command, drug_code=None):
"""Get information and summary statistics on a phenotype.
:param phenotype: The name of the phenotype to get summary information on.
:type phenotype: str
Use 'list' to see all the available phenotypes for this command.
This command can also be used to get information on a drug:
info drug 12345 or C05 (ATC code)
"""
if phen_or_command == "drug":
return _info_drug(drug_code)
phenotype = phen_or_command
message = StringIO()
data, meta = _get_data_meta(phenotype)
print("Phenotype meta data:", file=message)
for k, v in meta.items():
if COLOR:
k = colored(k, "green")
print("\t{}{}".format(k.ljust(30), v), file=message)
print("\nSummary statistics:", file=message)
n_missing = STATE["manager"].get_number_missing(phenotype)
n_total = data.shape[0]
print("\t{} / {} missing values ({:.3f}%)".format(
n_missing, n_total, n_missing / n_total * 100
), file=message)
t = types.type_str(meta["variable_type"])
if t.subtype_of(types.Discrete):
# Show information on prevalence.
n_cases = np.sum(data == 1)
n_controls = np.sum(data == 0)
print("\t{} cases, {} controls; prevalence: {:.3f}%".format(
n_cases, n_controls, n_cases / (n_cases + n_controls) * 100
), file=message)
elif t.subtype_of(types.Continuous):
mean = np.nanmean(data)
std = np.nanstd(data)
print(u"\tµ = {}, σ = {}".format(mean, std), file=message)
print("\tmin = {}, max = {}".format(np.nanmin(data), np.nanmax(data)),
file=message)
elif t.subtype_of(types.Factor):
print("\nCounts (rate):", file=message)
n = data.shape[0]
for name, count in data.value_counts().iteritems():
print("\t{}: {} ({:.3f}%)".format(name, count, count / n * 100),
file=message)
return {"success": True, "message": message.getvalue()}
def normal_qq_plot(phenotype):
"""Plot the Normal QQ plot of the observations.
:param phenotype: The phenotype for which to draw the QQ plot.
:type phenotype: str
This function is only available for continuous phenotypes.
"""
data, meta = _get_data_meta(phenotype)
data = data[~np.isnan(data)]
if types.type_str(meta["variable_type"]).subtype_of(types.Continuous):
raise REPLException(
"Could not create QQ plot for {} variable '{}'.".format(
meta["variable_type"], phenotype
)
)
data = np.sort(data)
expected = scipy.stats.norm.ppf(
np.arange(1, data.shape[0] + 1) / (data.shape[0] + 1),
loc=np.mean(data),
scale=np.std(data)
)
plt.scatter(expected, data, color="black", marker="o", s=10)
x_min, x_max = plt.xlim()
y_min, y_max = plt.ylim()
slope, intercept, r_value, p_value, std_err = scipy.stats.linregress(
expected, data
)
plt.plot(
[x_min, x_max],
[slope * x_min + intercept, slope * x_max + intercept],
"--", color="black",
label="$R^2 = {:.4f}$".format(r_value ** 2)
)
plt.legend(loc="lower right")
plt.xlabel("Expected quantiles")
plt.ylabel("Observed quantiles")
plt.xlim([x_min, x_max])
plt.ylim([y_min, y_max])
filename = "cohort_plot.png"
return _response_from_img_filename(filename)
def histogram(phenotype, nbins=None):
"""Draw a histogram (or a bar plot for discrete variables) of the data.
:param phenotype: The phenotype for which to draw the histogram.
:type phenotype: str
:param nbins: The number of bins for the histogram (optional).
:type nbins: int
This function will work on both continuous and discrete variables (but
not factors).
"""
data, meta = _get_data_meta(phenotype)
t = types.type_str(meta["variable_type"])
data = data[~np.isnan(data)]
if t.subtype_of(types.Continuous):
# Histogram.
if nbins:
plt.hist(data, bins=nbins)
else:
plt.hist(data)
plt.xlabel(phenotype)
elif t.subtype_of(types.Discrete):
# Bar plot.
plt.bar((0.1, 0.4), (np.sum(data == 0), np.sum(data == 1)), width=0.1)
plt.xticks((0.15, 0.45), ("control", "case"))
plt.xlim((0, 0.6))
else:
raise REPLException("Could not generate histogram for '{}' variable."
"".format(meta["variable_type"]))
filename = "cohort_plot.png"
plt.savefig(filename, dpi=300)
return _response_from_img_filename(filename)
