def _get_global_props_data_to_plot(cfg, weighted, props, load_stats=False):
"""
Internal helper function, for fetching data for global
network properties.
Returns
-------
propMeans : a python dictionary with keys as the property
propMeans[prop][mode][p] = mean
propMeanConfIntervals : dict
propData[prop][mode][p] = [low, high]
propStats : a python dict containing the perm. ttest pvalues
propStats[prop][p] = pvalue
densities : list
of the densities
"""
if weighted:
props_tag = 'global_w_props'
else:
props_tag = 'global_uw_props'
fnGroups = [cfg['group_1_mat_fnames'], cfg['group_2_mat_fnames']]
data = [
dataio.merge_and_load_props_data(fnGroup, props_tag, props, cfg)
for fnGroup in fnGroups
]
if load_stats:
statsData = dataio.load_pickle(fnc.get_stats_fname(cfg, props_tag))
propMeans = {}
propMeanConfIntervals = {}
propPVals = {}
densities = data[0][settings.densities_tag]
for prop in data[0].keys():
if prop in props:
modeMeans = []
modeMeanErrs = []
for i, datum in enumerate(data):
# if prop != settings.densities_tag:
propData = datum[prop]
modeMeans.append(np.average(propData, 0))
modeMeanErrs.append(
bootstrap.mean_conf_interval(propData.T,
cfg['bootstrap_samples'],
cfg['bootstrap_coverage']))
propMeans[prop] = np.array(modeMeans)
propMeanConfIntervals[prop] = np.array(modeMeanErrs)
if load_stats:
propPVals[prop] = statsData[prop][settings.pval_tag]
return propMeans, propMeanConfIntervals, propPVals, densities
def plot_link_sim_mat(cfg):
linkSimMatrix = dataio.load_pickle(
fnc.get_fname(cfg, settings.common_links_tag))
linkSimMatrix = linkSimMatrix / float(linkSimMatrix[0, 0]) # get to ratio
fig = plt.figure(figsize=(8, 10))
ax = fig.add_subplot(1, 1, 1)
measure = settings.common_links_tag
plot_sim_mat(ax, linkSimMatrix, measure)
_addPermutationTestAndTextsToSimMatrix(fig, linkSimMatrix, measure)
ax.set_xticklabels(2 * [str(i) for i in range(1, 14)])
ax.set_yticklabels(2 * [str(i) for i in range(1, 14)])
fig.savefig(cfg["outdata_dir"] + measure + ".pdf",
format="pdf",
bbox_inches='tight')
def plotDensityVsAvgPairedShare(cfg):
"""
Plots the avg. fraction of the common links between the two settings
(paired setup)
"""
data = dataio.load_pickle(fnc.get_fname(cfg, settings.common_links_tag))
linkSimMatrices = data[settings.common_links_tag]
n = len(linkSimMatrices[0]) / 2
densities = data[settings.densities_tag]
linkSimAvgs = []
linkSimStds = []
vals = []
for linkSimMatrix in linkSimMatrices:
linkSimMatrix = linkSimMatrix / \
float(linkSimMatrix[0, 0]) # get to ratio
linkSimMatrix[range(n)]
vals.append(linkSimMatrix[range(n), n + np.array(range(n))])
linkSimAvgs.append(np.average(vals[-1]))
linkSimStds.append(np.std(vals[-1]))
fig = plt.figure()
ax = fig.add_subplot(111)
densities = np.array(densities)
ax.set_xlabel(settings.get_prop_tex_name(settings.densities_tag))
ax.set_ylabel(r"Share of common links")
vals = np.array(vals).T
for i, val in enumerate(vals):
ax.plot(densities, val, "r", alpha=0.4, label=str(i + 1))
ax.errorbar(densities, linkSimAvgs, linkSimStds, color="b", label="avg")
ax.set_xscale('log')
l = ax.legend(loc=1)
l.get_frame().set_alpha(0.75)
if cfg['include_mst']:
plt.savefig(cfg["outdata_dir"] + settings.common_links_tag +
"_paired_vs_" + settings.densities_tag + "_with_mst.pdf",
format="pdf",
bbox_inches="tight")
else:
plt.savefig(cfg["outdata_dir"] + settings.common_links_tag +
"_paired_vs_" + settings.densities_tag + ".pdf",
format="pdf",
bbox_inches="tight")
def plotSameLinksShareVsLouvainSimilarityMeasures(cfg,
filenamesGroup1,
filenamesGroup2,
density=None):
allFNames = filenamesGroup1 + filenamesGroup2
data = dataio.mergeAndLoadLouvainProperties(allFNames, density)
clusterings = data[settings.louvain_cluster_tag]
simMatricesDict = gencomps.computeClusterSimilarityMeasures(clusterings)
linkSimMatData = dataio.load_pickle(
fnc.get_fname(cfg, settings.common_links_tag))
index = 0
densities = linkSimMatData[settings.densities_tag]
if density is None:
pass
else:
for j in range(len(densities)):
if int(densities[index]) == int(density):
index = j
break
linkSimMatrix = linkSimMatData[settings.common_links_tag][index]
linkSimMatrix = linkSimMatrix / float(linkSimMatrix[0, 0]) # get to ratio
triu_indices = np.triu_indices_from(linkSimMatrix, 1)
for measure, simMatrix in simMatricesDict.iteritems():
simMeasures = simMatrix[triu_indices]
linkSims = linkSimMatrix[triu_indices]
fig = plt.figure()
ax = fig.add_subplot(111)
ax.plot(linkSims, simMeasures, "o")
ax.set_ylabel(settings.get_prop_tex_name(measure))
ax.set_xlabel(settings.get_prop_tex_name(settings.common_links_tag))
(r, p) = pearsonr(linkSims, simMeasures)
fig.suptitle(r"Pearson correlation: {:.3f}".format(r))
fig.savefig(cfg["outdata_dir"] + settings.common_links_tag + "_vs_" +
measure + ".pdf",
format="pdf",
bbox_inches='tight')
def plotDensityVsAvgCommonFractionOfLinksOverAllPairsInDifferentGroups(cfg):
data = dataio.load_pickle(fnc.get_fname(cfg, settings.common_links_tag))
linkSimMatrices = data[settings.common_links_tag]
densities = data[settings.densities_tag]
linkSimAvgs = []
linkSimStds = []
n = len(linkSimMatrices[0]) / 2
differentGroupIndices = (np.array((np.ones(
(n, n)) - np.eye(n)).nonzero()).T + np.array([n, n])).T
print differentGroupIndices
for linkSimMatrix in linkSimMatrices:
linkSimMatrix = linkSimMatrix / \
float(linkSimMatrix[0, 0]) # get to ratio
linkSimAvgs.append(np.average(linkSimMatrix[differentGroupIndices]))
linkSimStds.append(np.std(linkSimMatrix[differentGroupIndices]))
fig = plt.figure()
ax = fig.add_subplot(111)
densities = np.array(densities)
ax.errorbar(densities,
linkSimAvgs,
linkSimStds,
label="avg. over all pairs in different groups, neglecting" +
" same subjects, stdev")
ax.set_xscale('log')
l = ax.legend(loc=1)
l.get_frame().set_alpha(0.75)
if cfg['include_mst']:
plt.savefig(cfg["outdata_dir"] + settings.common_links_tag + "_vs_" +
settings.densities_tag + "_between_groups.pdf",
format="pdf",
bbox_inches="tight")
else:
plt.savefig(cfg["outdata_dir"] + settings.common_links_tag + "_vs_" +
settings.densities_tag + "_between_groups_noMST.pdf",
format="pdf",
bbox_inches="tight")
consensus_fname1 = (cfg['outdata_dir'] + comdet_tag + "_" + "group2" + "_" +
str(cfg['density']) + ".pkl")
compcoms.comp_consensus_partition(cfg,
'group_1_mat_fnames',
consensus_fname0,
n_to_consider=n_to_consider,
comdet_tag=comdet_tag)
compcoms.comp_consensus_partition(cfg,
'group_2_mat_fnames',
consensus_fname1,
n_to_consider=n_to_consider,
comdet_tag=comdet_tag)
print "Matching modules algorithmically"
consensus_clu_movie = \
dataio.load_pickle(consensus_fname0)[comdet_tag]
consensus_clu_rest = \
dataio.load_pickle(consensus_fname1)[comdet_tag]
consensus_clu_movie, consensus_clu_rest, n_clu_1, n_clu_2 = \
viz._get_matched_renumbered_communities_for_viz(
consensus_clu_movie, consensus_clu_rest
)
uf_consensus_clus = [consensus_clu_movie, consensus_clu_rest]
print "Setting colors for modules"
ok_nodes = dataio.get_ok_nodes(cfg['blacklist_fname'])
filtered_consensus_clus = [
consensus_clu_movie[ok_nodes], consensus_clu_rest[ok_nodes]
]
def plot_link_dist_probs_by_condition(cfg):
"""
Plots the link distance PDFs pooled by condition.
Parameters
----------
cfg : dict
the user-specified config dictionary.
The following keys are required::
"group_1_mat_fnames"
"group_2_mat_fnames"
"group_1_color"
"group_2_color"
"group_1_label"
"group_2_label"
"outdata_dir"
Returns
-------
None
"""
# CFGCHANGE?
config.require(cfg, [
"group_1_mat_fnames", "group_2_mat_fnames", "group_1_color",
"group_2_color", "group_1_label", "group_2_label", "outdata_dir",
"density"
])
fname_group_list = [cfg["group_1_mat_fnames"], cfg["group_2_mat_fnames"]]
colors = [cfg["group_1_color"], cfg["group_2_color"]]
labels = [cfg["group_1_label"], cfg["group_2_label"]]
# ns = []
data_example = dataio.load_pickle(
fnc.get_ind_fname(fname_group_list[0][0], cfg,
settings.link_distance_tag))
densities = data_example[settings.config_tag][settings.densities_tag]
# for individual plots:
# figs = [plt.figure() for p in densities]
# axs = [fig.add_subplot(111) for fig in figs]
group_distances = []
for i, fname in enumerate(fname_group_list):
# ns.append(len(fname))
distances = [np.array([]) for p in densities]
for fname in fname:
print fname
data = dataio.load_pickle(
fnc.get_ind_fname(fname, cfg, settings.link_distance_tag))
p_f_dists = data[settings.link_distance_tag]
for j, p in enumerate(densities):
# print p
distances[j] = np.hstack((distances[j], p_f_dists[j]))
# for individual plots:
# for j in range(len(densities)):
# genplots.plot_inv_cdf(
# axs[j], distances[j], yscale='log',
# label=labels[i], color=colors[i])
group_distances.append(distances)
indices = range(len(densities)) # [6,7,10] #density indices
print densities
for k, j in enumerate(indices):
fig = plt.figure(figsize=(4, 3))
p = densities[j]
ax = fig.add_subplot(1, 1, 1)
ax.set_xlabel(settings.get_prop_tex_name(settings.link_distance_tag))
ax.set_ylabel(r"1-CDF(d)")
print p, j
ax.text(.5,
0.10,
r"$\rho$ = " + str(p * 100) + "\%",
ha='center',
va='center',
transform=ax.transAxes)
for i in range(len(fname_group_list)):
genplots.plot_inv_cdf(ax,
group_distances[i][j],
label=labels[i],
color=colors[i],
yscale='log')
plt.tight_layout()
fig.savefig(cfg['outdata_dir'] + "linkDistProbs_" + str(p) + ".pdf",
format="pdf",
bbox_inches="tight")
plt.close(fig)
return None