def plot_scatters(subjects, axes):
ftemp = "correlation_analysis/{}_{}_ifs.pkz"
for subj, ax in zip(subjects, axes):
sticks = moss.load_pkl(ftemp.format(subj, "sticks")).corrmat
rest = moss.load_pkl(ftemp.format(subj, "rest")).corrmat
triu = np.triu_indices_from(rest, 1)
ax.scatter(sticks[triu],
rest[triu],
s=3,
linewidth=.2,
color=".6",
edgecolor="w",
rasterized=True)
ax.plot([-.2, .8], [-.2, .8], lw=1, dashes=(5, 2), color=".3")
plt.setp(axes,
xlim=(-.25, .8),
ylim=(-.25, .8),
xticks=np.linspace(-.2, .8, 6),
yticks=np.linspace(-.2, .8, 6),
aspect="equal")
plt.setp(axes[1:], yticklabels=[])
for ax in axes:
sns.despine(ax=ax, trim=True)
plt.setp(ax.get_xticklabels(), size=6)
plt.setp(ax.get_yticklabels(), size=6)
def plot_kdes(subjects, axes):
ftemp = "correlation_analysis/{}_{}_ifs.pkz"
for subj, ax in zip(subjects, axes):
sticks = moss.load_pkl(ftemp.format(subj, "sticks")).corrmat
rest = moss.load_pkl(ftemp.format(subj, "rest")).corrmat
triu = np.triu_indices_from(rest, 1)
sns.kdeplot(sticks[triu], color=".15", label="residual", ax=ax)
sns.kdeplot(rest[triu],
color=".45",
dashes=[4, 1],
label="resting",
ax=ax)
plt.setp(axes,
xlim=(-.25, .8),
ylim=(0, 17),
xticks=np.linspace(-.2, .8, 6),
yticks=[])
for ax in axes:
sns.despine(ax=ax, left=True, trim=True)
plt.setp(ax.get_xticklabels(), size=6)
plt.setp(ax.get_yticklabels(), size=6)
axes[0].legend(bbox_to_anchor=(1.2, .8))
for ax in axes[1:]:
ax.legend_ = None
def plot_time_corrs(subjects, axes):
x = np.arange(1, 5)
palette = [".2", ".5"]
for subj, ax in zip(subjects, axes):
res_fname = "correlation_analysis/{}_rest_ifs.pkz".format(subj)
res = moss.load_pkl(res_fname)
for line, color in zip(res.corr_times.T, palette):
ax.plot(x, line, "o-", color=color, ms=3, clip_on=False)
sig = res.corr_times_pctiles > 95
ax.plot(x[sig],
np.ones(sig.sum()) * .0025,
marker=(6, 2, 0),
ls="",
mew=.35,
mec=".2",
ms=3)
ax.set(xticks=x, xlim=(.6, 4.4), ylim=(0, .07))
sns.despine(ax=ax, trim=True)
plt.setp(axes[1:], yticklabels=[])
axes[0].set_ylabel("Correlation (r)")
def plot_mds(subjects, experiments, axes):
for subj, exp, ax in zip(subjects, experiments, axes):
res_fname = "correlation_analysis/{}_{}_ifs.pkz".format(subj, exp)
res = moss.load_pkl(res_fname)
sorter = np.argsort(np.abs(res.prefs))
x_, y_ = res.mds_coords.T.dot(res.prefs)
t = np.arctan2(y_, x_)
rot = [[np.cos(t), np.sin(t)], [-np.sin(t), np.cos(t)]]
x, y = np.dot(rot, res.mds_coords[sorter].T)
cmap = get_colormap(exp)
ax.scatter(x,
y,
c=res.prefs[sorter],
cmap=cmap,
vmin=-1.75,
vmax=1.75,
s=8,
linewidth=0)
ax.set(xlim=(-.9, .9), ylim=(-.9, .9), aspect="equal")
ax.set_axis_off()
def get_subject_order(exp):
subjects = lyman.determine_subjects([exp + "_subjects"])
accs = pd.Series(index=subjects, dtype=np.float)
for subj in subjects:
fname = "decoding_analysis/{}_{}_ifs.pkz".format(subj, exp)
accs.ix[subj] = moss.load_pkl(fname).acc
return list(accs.sort(inplace=False, ascending=False).index)
def plot_corrmats(subjects, axes, exp):
for subj, ax in zip(subjects, axes):
fname = "correlation_analysis/{}_{}_ifs.pkz".format(subj, exp)
corrmat = moss.load_pkl(fname).corrmat
ax.imshow(corrmat - np.eye(len(corrmat)),
cmap="RdBu_r",
vmin=-.15,
vmax=.15,
rasterized=True)
ax.set(xticks=[], yticks=[])
sns.despine(ax=ax, left=True, bottom=True)
def plot_prediction_curves(subjects, axes, exp):
res_ftemp = "spatial_analysis/{}_{}_ifs.pkz"
for subj, ax in zip(subjects, axes):
res = moss.load_pkl(res_ftemp.format(subj, exp))
x = res.steps
norm = res.null.mean()
real = res.real / norm
pint = res.pint / norm
ax.plot(x, real, "o-", color=".15", ms=2.5, clip_on=False)
ax.fill_between(x, *pint, color=".4", alpha=.3)
cross_x, cross_y = res.intersect
cross_y /= norm
ax.plot([cross_x, cross_x], [0, cross_y],
lw=.8,
dashes=[3, 1],
color=".5",
zorder=0)
ax.set(xlim=(0, 40),
ylim=(0, 2),
xticks=np.linspace(0, 40, 5),
yticks=[0, 1, 2],
yticklabels=[0, 1, 2])
sns.despine(ax=ax)
ylabel = "Normalized error"
plt.setp(axes[1:7], yticklabels=[])
axes[0].set(ylabel=ylabel)
if exp == "dots":
plt.setp(axes[8:], yticklabels=[])
plt.setp(axes[:7], xticklabels=[])
axes[7].set_ylabel(ylabel)
def plot_brains(subjects, axes):
for subj, subj_axes in zip(subjects, axes):
exp = dict(pc="dots", ti="sticks")[subj[:2]]
data_fname = "roi_cache/{}_{}_ifs.npz".format(subj, exp)
with np.load(data_fname) as dobj:
vox_ijk = dobj["vox_ijk"]
res_fname = "decoding_analysis/{}_{}_ifs.pkz".format(subj, exp)
res = moss.load_pkl(res_fname)
prefs = res.prefs
surf_vals = roi_to_surf(exp, subj, prefs, vox_ijk)
lut = get_colormap(exp, False)
for hemi, ax in zip(["lh", "rh"], subj_axes):
b = Brain(subj,
hemi,
"inflated",
background="white",
cortex=("binary", -4, 8, False),
size=(1000, 600))
b.add_data(surf_vals.ix[hemi].fillna(-11).values,
colormap=lut,
colorbar=False,
thresh=-10,
min=-1.75,
max=1.75)
mlab.view(*get_ifs_view(subj, hemi))
img = crop(b.screenshot())
ax.imshow(img, rasterized=True)
ax.set(xticks=[], yticks=[])
b.close()
def plot_cluster_error(ax):
res_ftemp = "spatial_analysis/{}_{}_ifs.pkz"
for exp in ["dots", "sticks"]:
subjects = get_subject_order(exp)
color = get_colormap(exp, as_cmap=False)[20]
errs = []
for subj in subjects:
res = moss.load_pkl(res_ftemp.format(subj, exp))
x = res.steps
norm = res.null.mean()
errs.append(res.real / norm)
errs = np.vstack(errs)
mean = errs.mean(axis=0)
ax.plot(x, mean, color=color, lw=2)
sem = stats.sem(errs, axis=0)
ax.fill_between(x, mean - sem, mean + sem, alpha=.2, color=color)
ax.axhline(y=1,
lw=1,
dashes=[5, 2],
color=".5",
zorder=0,
xmin=.02,
xmax=.98)
ax.set(xlim=(0, 42),
ylim=(.55, 1.45),
yticks=[.6, .8, 1, 1.2, 1.4],
xticks=[0, 10, 20, 30, 40],
xlabel="Neighborhood radius (mm)",
ylabel="Normalized error")
sns.despine(ax=ax, trim=True)
def plot_hists(subjects, axes, label_last=1, ymax=350):
bins = np.linspace(-2, 2, 20)
for subj, ax in zip(subjects, axes):
exp = dict(pc="dots", ti="sticks")[subj[:2]]
res_fname = "decoding_analysis/{}_{}_ifs.pkz".format(subj, exp)
res = moss.load_pkl(res_fname)
prefs = res.prefs
pctiles = res.pref_pctiles
cmap = get_colormap(exp)
plot_prefs = [
prefs[pctiles < 10], prefs[pctiles > 90],
prefs[(pctiles >= 10) & (pctiles <= 90)]
]
ax.hist(plot_prefs,
histtype="barstacked",
rwidth=1,
color=[cmap(.01), cmap(.99), ".9"],
bins=bins)
ax.set(xlim=(-2, 2),
xticks=[-2, -1, 0, 1, 2],
xticklabels=[],
yticks=[],
ylim=(0, ymax))
for ax in axes[-label_last:]:
ax.set_xlabel("Context\npreference", labelpad=2, fontsize=7)
ax.set(xticklabels=[-2, -1, 0, 1, 2])
for ax in axes:
sns.despine(ax=ax, left=True)
def plot_distance_corrs(subjects, axes, exp):
for subj, ax in zip(subjects, axes):
res_fname = "correlation_analysis/{}_{}_ifs.pkz".format(subj, exp)
res = moss.load_pkl(res_fname)
x = res.distance_thresh
for dim, color, marker in zip(["3D", "2D"], [".5", ".2"], ["x", "+"]):
same, diff = res.corr_distance[dim].T
ax.plot(x, same - diff, "o-", color=color, ms=3, clip_on=False)
sig = res.corr_distance_pctiles[dim] > 95
stary = -.005 if exp == "dots" else -.0025
ax.plot(x[sig],
np.ones(sig.sum()) * stary,
marker=marker,
ls="",
mew=.35,
mec=".2",
ms=3)
ylim = (-.01, .08) if exp == "dots" else (-.005, .04)
yticks = np.array([0, .01, .02, .03, .04])
yticks = yticks * 2 if exp == "dots" else yticks
ax.set(xlim=(-2, 42), ylim=ylim, yticks=yticks)
sns.despine(ax=ax, trim=True)
ylabel = "Subnetwork strength\n($r_{\mathrm{same}} - r_{\mathrm{diff}}$)"
plt.setp(axes[1:7], yticklabels=[])
axes[0].set_ylabel(ylabel)
if exp == "dots":
plt.setp(axes[8:], yticklabels=[])
plt.setp(axes[:7], xticklabels=[])
axes[7].set_ylabel(ylabel)
rois = ["ifs", "mfc"]
decoding_data = {}
for exp, subj_list in subjects.iteritems():
# Set up the dataframe for this experiment
idx = pd.MultiIndex.from_product([subj_list, rois],
names=["subj", "roi"])
exp_df = pd.DataFrame(index=idx,
columns=["acc", "chance", "pctile"],
dtype=np.float)
# Load the data for each subject/roi
for subj, roi in idx:
fname = "decoding_analysis/{}_{}_{}.pkz".format(subj, exp, roi)
res = moss.load_pkl(fname)
exp_df.ix[subj, roi] = [res.acc, res.chance, res.acc_pctile]
decoding_data[exp] = exp_df
# Combine across experiments and save
decoding_df = pd.concat(decoding_data, names=["experiment"]).reset_index()
decoding_df.to_csv("data/decoding_results.csv", index=False)
# ----- Combine correlation results
subjects["rest"] = subjects["sticks"]
corr_dfs = {}
for exp in ["dots", "sticks", "rest"]:
subj_list = subjects[exp]
exp_data = []
if __name__ == "__main__":
try:
_, subj, exp, roi = sys.argv
except ValueError:
sys.exit("Usages: spatial_analysis.py <subj> <exp> <roi>")
# Ensure that the output exists
if not op.exists("spatial_analysis"):
os.mkdir("spatial_analysis")
# Load the context preference data
fname = "decoding_analysis/{}_{}_{}.pkz".format(subj, exp, roi)
prefs = moss.load_pkl(fname).prefs
# Load the distance matrix
fname = "roi_cache/{}_{}_{}.npz".format(subj, exp, roi)
with np.load(fname) as dobj:
dmat = dobj["dmat2d"]
# Define the steps and radius
steps = np.arange(2, 42, 2)
radius = 2
# Compute the real curve
real = prediction_curve(dmat, prefs, steps, radius)
# Compute the null curves
seed = sum(map(ord, subj + "_spatial"))
sys.exit("Usages: decoding_analysis.py <subj> <exp> <roi>")
# Ensure that the output exists
if not op.exists("correlation_analysis"):
os.mkdir("correlation_analysis")
# Load the data
data_fname = "roi_cache/{}_{}_{}.npz".format(subj, exp, roi)
dobj = np.load(data_fname)
data = dobj["ts_data"]
dmats = {"2D": dobj["dmat2d"], "3D": dobj["dmat3d"]}
# Load the outputs of the decoding analysis
task_exp = "sticks" if exp == "rest" else exp
res_fname = "decoding_analysis/{}_{}_{}.pkz".format(subj, task_exp, roi)
res = moss.load_pkl(res_fname)
# Remove voxels that were excluded from the decoding analysis
data = data[:, res.good_voxels]
# Pull out the tail voxels
tail_mask = res.tails.astype(np.bool)
data = data[:, tail_mask]
tails = res.tails[tail_mask]
prefs = res.prefs[tail_mask]
# Regress the task-related effects out of experiment data
if exp != "rest":
data = regress_task(exp, subj, data)
# Compute the timeseries correlation