def _report_features(features0, features1, features, NC):
if logging.getLogger("output").getEffectiveLevel() > logging.DEBUG: return
plots.pyplot_reset()
logging.debug(
"[sample_data] Storing features to /tmp/sample_data_features.tsv")
features.to_csv("/tmp/sample_data_features.tsv",
sep="\t",
header=True,
index=False)
if NC < 2: return
x = np.asarray(features0[["cov0", "cov1"]])
z = np.asarray(list(zip(1 - features0["group"], features0["group"])))
_plot_data_2d(x, z, label="$i_u=0$")
x = np.asarray(features1[["cov0", "cov1"]])
z = np.asarray(list(zip(1 - features1["group"], features1["group"])))
_plot_data_2d(x, z, label="$i_u=1$")
pyplot.xlabel("covariate 0")
pyplot.ylabel("covariate 1")
pyplot.grid(True)
pyplot.legend()
plots.savefig("/tmp/sample_data_features.png")
data["St"], lower, upper = theoretic_pvalues_rejections(
ldf, ratios, "SHfH0_pval", badge_time)
data["Se"], lower, upper = empirical_pvalues_rejections(
ldf, ratios, "SHfH0_LLR", badge_time)
data["Be"], lower, upper = empirical_pvalues_rejections(
ldf, ratios, "BHfH0_LLR", badge_time)
plots.pyplot_reset()
plot_data(
data,
ylabel=r"$H_0$ rejection probability",
#level=0.05, level_label="5\%",
columns=["St", "Se", "Be"],
labels=["basic theoretic", "basic bootstrap", "robust bootrsap"])
pyplot.legend(fontsize=20, loc=4)
plots.savefig("%s_l%g_rejections.pdf" % (args.input, l0))
#####################################################################
##############################################################################################
print("Empirical vs theoretic p-values")
for l0, ldf in df.groupby("l0"):
#####################################################################
data = pd.DataFrame({"ratio": ratios})
data["St"], lower, upper = theoretic_pvalues(ldf, ratios, "SHfH0_pval",
badge_time)
data["Se"], lower, upper = empirical_pvalues(ldf, ratios, "SHfH0_LLR",
badge_time)
data["Be"], lower, upper = empirical_pvalues(ldf, ratios, "BHfH0_LLR",
badge_time)
data = pd.DataFrame({"ratio": ratios})
data["St"], lower, upper = theoretic_pvalues_rejections(ldf, ratios, "SHfH0_pval", badge_time)
data["Se"], lower, upper = empirical_pvalues_rejections(ldf, ratios, "SHfH0_LLR", badge_time)
data["Be"], lower, upper = empirical_pvalues_rejections(ldf, ratios, "BHfH0_LLR", badge_time)
plots.pyplot_reset()
pyplot.ylim((0, 1.0))
plot_data(data, ylabel=r"$H_0$ rejection probability",
#level=0.05, level_label="5\%",
columns=["St", "Se", "Be"],
labels=["basic theoretic", "basic bootstrap", "robust bootrsap"],
leg_loc=2)
pyplot.ylim((0, 1.0))
pyplot.gcf().subplots_adjust(bottom=0.17, left=0.18)
plots.savefig("%s_l%g_rejections_survival.pdf" % (args.input, l0))
#####################################################################
##############################################################################################
print("Empirical vs theoretic p-values")
for l0, ldf in df.groupby("l0"):
#####################################################################
data = pd.DataFrame({"ratio": ratios})
data["St"], lower, upper = theoretic_pvalues(ldf, ratios, "SHfH0_pval", badge_time)
data["Se"], lower, upper = empirical_pvalues(ldf, ratios, "SHfH0_LLR", badge_time)
data["Be"], lower, upper = empirical_pvalues(ldf, ratios, "BHfH0_LLR", badge_time)
plots.pyplot_reset()
pyplot.ylim((0, 1.0))
plot_data(data, ylabel=r"$p$-value", level=0.05,
lw=3,
color=COLORS[i % len(COLORS)])
#pyplot.plot(data[TIMECOL], means+stds, color=p[-1].get_color(), lw=1)
pyplot.ylabel("average intensity (days)", fontsize=25)
plots.pyplot_parse_params2(xmin=params.get("xmin", None),
xmax=params.get("xmax", None),
ymax=params.pop("ymax", None))
pyplot.grid(True)
_plot_badges(args)
xmin, xmax = pyplot.xlim()
plot_legend(fontsize=20,
loc=(1 if args.badges[0] < (xmin + xmax) + 0.5 else 2))
_set_time_axis(params)
pyplot.tick_params(axis='both', which='major', labelsize=22)
pyplot.gcf().subplots_adjust(bottom=0.17, left=0.22)
plots.savefig(args.output + "_fitting.pdf")
#########################################################################
print("=================================")
print("LLR-values over time")
plots.pyplot_reset()
#_plot_badges2(args)
#transform = lambda c: list(c.apply(lambda v: numpy.exp(v)))
transform = lambda c: list(c)
x, y = _smoothing(data[TIMECOL], transform(data["SHfH0_LLR"]), args,
params)
pyplot.plot(list(x), y, label="basic", lw=3, ls="-", color=COLORS[1])
#pyplot.plot(x[5::15], y[5::15], marker="o", markeredgecolor="none", markersize=5, color=COLORS[1], lw=0)
for trend, tdf in df.groupby("trend"):
for F, sdf in tdf.groupby("F"):
for col in VALUE_COLS:
logging.info("PROCESSING: %s" % col)
cmap = "Blues" if "TR" in col else "Reds"
m, rl, cl = extract_matrix(sdf,
column=col,
aggregate=np.mean,
row_column="ishift",
col_column="covmshift")
matplotlib.rcParams.update({'font.size': 24})
matplotlib.rcParams['pdf.fonttype'] = 42
matplotlib.rcParams['ps.fonttype'] = 42
matplotlib.rcParams['text.usetex'] = True
plot_matrix(m,
rl,
cl,
xlabel=r"covariates discrepancy, $\Delta_x$",
ylabel=r"badge effect, $\Delta_\lambda$",
clabel="AUC",
cmin=0.5,
cmax=1.0,
cmap=cmap)
matplotlib.rcParams['pdf.fonttype'] = 42
matplotlib.rcParams['ps.fonttype'] = 42
matplotlib.rcParams['text.usetex'] = True
plots.savefig(output + "_t%s_F%s_%s.pdf" % (trend, F, col))