def do_simulations(params):
rt = []
response = []
stimulus = []
for stim in [1, 0]:
# get traces:
x = get_OU_traces(
v=params['v'],
ll=params['ll'],
dc=params['dc'],
z=params['z'],
pre_generated=False,
stim=stim,
nr_trials=params['nr_trials'],
tmax=tmax,
dt=dt,
)
# get bounds:
if params['bound'] == 'default':
b1, b0 = _bounds(a=params['a'], lower_is_0=False, tmax=tmax, dt=dt)
elif params['bound'] == 'collapse_linear':
b1, b0 = _bounds_collapse_linear(a=params['a'],
c1=params['c1'],
c0=params['c0'],
lower_is_0=False,
tmax=tmax,
dt=dt)
elif params['bound'] == 'collapse_hyperbolic':
b1, b0 = _bounds_collapse_hyperbolic(a=params['a'],
c=params['c'],
lower_is_0=False,
tmax=tmax,
dt=dt)
# apply bounds:
rt_dum, response_dum = apply_bounds_diff_trace(x=x, b1=b1, b0=b0)
# store results:
rt.append((rt_dum * dt) + ndt)
response.append(response_dum)
stimulus.append(np.ones(params['nr_trials']) * stim)
df = pd.DataFrame()
df.loc[:, 'rt'] = np.concatenate(rt)
df.loc[:, 'response'] = np.concatenate(response)
df.loc[:, 'stimulus'] = np.concatenate(stimulus)
df.loc[:, 'correct'] = np.array(
np.concatenate(stimulus) == np.concatenate(response), dtype=int)
df.loc[:, 'subj_idx'] = params['subj_idx']
df.to_csv(os.path.join(data_folder,
'df_{}.csv'.format(params['subj_idx'])))
def do_simulations(params):
# SIMULATE A SLOWLY AUTOCORRELATED DRIFT BIAS
v_autocorr = sample_signal(params['nr_trials'],
params['v_autocorr'],
mu=params['v'],
sigma=1)
dc_autocorr = sample_signal(params['nr_trials'],
params['dc_autocorr'],
mu=params['dc'],
sigma=1)
# SIMULATE A SEQUENCE OF RANDOM STIMULI
stims = np.repeat([0, 1], int(params['nr_trials'] / 2))
np.random.shuffle(stims)
## NOW DRAW RESPONSES FROM THAT PROCESS!
rt = []
response = []
stimulus = []
for t in tqdm(range(params['nr_trials'])):
# get traces:
x = get_DDM_traces(
v=v_autocorr[t],
z=params['z'],
dc=dc_autocorr[t],
dc_slope=0,
sv=params['sv'],
stim=stims[t],
nr_trials=1,
tmax=tmax,
dt=dt,
)
# get bounds:
if params['bound'] == 'default':
b1, b0 = _bounds(a=params['a'], tmax=tmax, dt=dt)
elif params['bound'] == 'collapse_linear':
b1, b0 = _bounds_collapse_linear(a=params['a'],
c1=params['c1'],
c0=params['c0'],
tmax=tmax,
dt=dt)
elif params['bound'] == 'collapse_hyperbolic':
b1, b0 = _bounds_collapse_hyperbolic(a=params['a'],
c=params['c'],
tmax=tmax,
dt=dt)
# apply bounds:
rt_dum, response_dum = apply_bounds_diff_trace(x=x, b1=b1, b0=b0)
# store results for single rtial
rt.append((rt_dum * dt) + ndt)
response.append(response_dum)
stimulus.append(stims[t])
df = pd.DataFrame()
df.loc[:, 'rt'] = np.concatenate(rt)
df.loc[:, 'response'] = np.concatenate(response)
df.loc[:, 'stimulus'] = stimulus
df.loc[:, 'correct'] = np.array(stimulus == np.concatenate(response),
dtype=int)
df.loc[:, 'subj_idx'] = params['subj_idx']
# NOW RECODE REPEATS/ALTERNATES INTO RESPONSES
df['prevresp'] = np.roll(df['response'], 1)
df['repeat'] = 1 * (df['prevresp'] == df['response'])
df['choice'] = df['response'].copy()
df['response'] = df['repeat']
df.to_csv(os.path.join(data_folder,
'df_{}.csv'.format(params['subj_idx'])))
list(np.arange(23, 34)),
]
for i, group in enumerate(groups):
df = pd.concat([
pd.read_csv(os.path.join(data_folder, 'df_{}.csv'.format(g)))
for g in group
],
axis=0)
df.to_csv(os.path.join(fits_folder, '2018_ou_data_{}.csv'.format(i + 1)))
# hyperbolic collapse:
fig = plt.figure()
t = np.arange(0, tmax, dt)
b1, b0 = _bounds_collapse_hyperbolic(a=a,
c=c,
lower_is_0=False,
tmax=tmax,
dt=dt)
plt.plot(t, b1)
plt.plot(t, b0)
fig.savefig(os.path.join(fig_folder, 'collapse_hyperbolic.pdf'))
# load ddm results:
for i, group in enumerate(groups):
# simulated data:
df = pd.read_csv(
os.path.join(fits_folder, '2018_ou_data_{}.csv'.format(i + 1)))
# model params:
params = []
def do_simulations(params):
rt = []
response = []
stimulus = []
traces = []
for stim in [1, 0]:
# get traces:
x = get_DDM_traces(
v=params['v'],
z=params['z'],
dc=params['dc'],
dc_slope=params['dc_slope'],
sv=params['sv'],
stim=stim,
nr_trials=params['nr_trials'],
tmax=tmax,
dt=dt,
)
# get bounds:
if params['bound'] == 'default':
b1, b0 = _bounds(a=params['a'], tmax=tmax, dt=dt)
elif params['bound'] == 'collapse_linear':
b1, b0 = _bounds_collapse_linear(a=params['a'],
c1=params['c1'],
c0=params['c0'],
tmax=tmax,
dt=dt)
elif params['bound'] == 'collapse_hyperbolic':
b1, b0 = _bounds_collapse_hyperbolic(a=params['a'],
c=params['c'],
tmax=tmax,
dt=dt)
# apply bounds:
rt_dum, response_dum = apply_bounds_diff_trace(x=x, b1=b1, b0=b0)
# store results:
rt.append((rt_dum * dt) + ndt)
response.append(response_dum)
stimulus.append(np.ones(params['nr_trials']) * stim)
traces.append(x)
df = pd.DataFrame()
df.loc[:, 'rt'] = np.concatenate(rt)
df.loc[:, 'response'] = np.concatenate(response)
df.loc[:, 'stimulus'] = np.concatenate(stimulus)
df.loc[:, 'correct'] = np.array(
np.concatenate(stimulus) == np.concatenate(response), dtype=int)
df.loc[:, 'subj_idx'] = params['subj_idx']
df.to_csv(os.path.join(data_folder,
'df_{}.csv'.format(params['subj_idx'])))
traces = np.vstack(traces)
for i in range(traces.shape[0]):
if sum(traces[i, :] > params['a']) > 0:
traces[i,
np.where(traces[i, :] > params['a'])[0][0]:] = params['a']
if sum(traces[i, :] < 0) > 0:
traces[i, np.where(traces[i, :] < 0)[0][0]:] = 0
hit = np.array((df['stimulus'] == 1) & (df['response'] == 1))
fa = np.array((df['stimulus'] == 0) & (df['response'] == 1))
miss = np.array((df['stimulus'] == 1) & (df['response'] == 0))
cr = np.array((df['stimulus'] == 0) & (df['response'] == 0))
shell()
fig = plt.figure(figsize=(2, 2))
for t in traces[hit, :][0:5000]:
plt.plot(t, alpha=0.005, lw=1, color='black')
# for t in traces[fa,:][0:500]:
# plt.plot(t, alpha=0.02, color='orange')
# for t in traces[miss,:][0:500]:
# plt.plot(t, alpha=0.02, color='green')
# for t in traces[cr,:][0:500]:
# plt.plot(t, alpha=0.02, color='green')
for trial, color, alpha in zip([hit, fa, miss, cr],
['orange', 'orange', 'green', 'green'],
[1, 0.5, 0.5, 1]):
y = np.nanmean(traces[trial, :], axis=0)
x = np.arange(y.shape[0])
ind = np.zeros(y.shape[0], dtype=bool)
ind[0:20] = True
(m, b) = sp.polyfit(x[ind], y[ind], 1)
regression_line = sp.polyval([m, b], x)
plt.plot(y, color=color, lw=2, alpha=alpha)
plt.plot(x[ind], regression_line[ind], color='black', lw=1, ls='--')
print(m)
plt.axhline(0, color='k', lw=0.5)
plt.axhline(params['a'], color='k', lw=0.5)
plt.axhline(params['z'], color='k', lw=0.5)
plt.xlim(0, 30)
plt.ylim(-0.1, params['a'] + 0.1)
plt.xlabel('Timesteps')
plt.ylabel('Decision variable')
sns.despine()
plt.tight_layout()
fig.savefig(
os.path.join(fig_folder,
'simulate_slopes_{}.pdf'.format(params['subj_idx'])))