'site': site,
'probe': probe,
'meta': meta
},
classobj_name=dPCA_anal_name,
cache_folder=
f'/home/mateo/mycache/{analysis_name}/{analysis_parameters}')
# dPCA_real, dPCA_shuffled, dPCA_simulated = get_cache(dPCA_anal)
except:
bad_sites.append(f"{site}_P{probe}_dPCA")
continue
for transformation, cache in zip(['LDA', 'dPCA'], [LDA_anal, dPCA_anal]):
real, shuffled, simulated = get_cache(cache)
for montecarlo, MCarray in zip(
['context discrimination', 'population effect'],
[shuffled, simulated]):
# calculates a signed pvalue, the signs is indicative of the direction, with possitive being higher and
# negative being lower than the mean of the Montecarlo distribution. Bi virtue of this distinction
# the p value is calculated on a single tail.
mont_mean = np.mean(MCarray, axis=0)
neg_pval = np.sum((MCarray < real), axis=0) / meta['montecarlo']
pos_pval = np.sum((MCarray > real), axis=0) / meta['montecarlo']
pvalues = np.where(real >= mont_mean, pos_pval, -neg_pval)
'probe_names': [probe],
'context_transitions': trans_pair,
'probe_order': [probe],
'trans_order': meta['transitions'],
'shuffle_num': meta['nonparam_shuffle'],
'trial_combinations': True
}
shuffled_dispersion_time = make_cache(
function=ndisp.transition_pair_comparison_by_trials,
func_args=func_args,
classobj_name=signal_name,
recache=False,
cache_folder=
f'/home/mateo/mycache/{analysis_name}/{analysis_parameters}')
real, trans_shuffle, trial_shuffle = get_cache(
shuffled_dispersion_time)
# pvalue = (msd_floor > obs_msd).sum() / shuffle_n
pvalues = np.sum(
(real < trans_shuffle), axis=0) / meta['nonparam_shuffle']
for pval_threshold in [0.05, 0.01, 0.001]:
signif = pvalues < pval_threshold
total_sig = np.sum(signif)
try:
last_sig = np.max(np.argwhere(signif))
except: # if there is not significant
last_sig = 0
d = {
'site': site,
nway = make_cache(
function=dPrime_from_NIT_site,
func_args={
'site': site,
'duration': duration,
'source': source,
'position': position,
'meta': meta
},
classobj_name=fourway_name,
cache_folder=
f'/home/mateo/mycache/{analysis_name}/{analysis_parameters}',
recache=False)
real, shuffled, simulated, transitions, contexts = get_cache(
nway) # get_cache(twoway)
if real is None:
continue
trans_pairs = itt.combinations(transitions, 2)
trans_pairs = [sorted(tp) for tp in trans_pairs
] # makes sure to not have repeated pairs as a-b b-a
for tt, trans_pair in enumerate(trans_pairs):
pair_name = f'{trans_pair[0]}_{trans_pair[1]}'
all_real[pair_name].append(real[tt, :])
all_shuffled[pair_name].append(shuffled[:, tt, :])
all_simulated[pair_name].append(simulated[:, tt, :])
# stacs all the individual probe instances across a new axis
'site': site,
'probe': probe,
'meta': meta
},
classobj_name=fourway_name,
cache_folder=
f'/home/mateo/mycache/{analysis_name}/{analysis_parameters}')
# twoway_name = f'191014_{site}_P{probe}_twoway_analysis'
#
# twoway = make_cache(function=twoway_analysis,
# func_args={'site': site, 'probe': probe, 'meta': meta},
# classobj_name=twoway_name,
# cache_folder=f'/home/mateo/mycache/{analysis_name}/{analysis_parameters}')
four = get_cache(fourway) # get_cache(twoway)
for analysis, a_name in zip([four], ['fourway']):
real, shuffled, simulated = analysis
# pairwise dprime
trans_pairs = list(itt.combinations(meta['transitions'], 2))
trans_count = 0
nrow = 2
ncol = 3
fig, axes = plt.subplots(nrow,
ncol,
sharey=True,
sharex=True,
sites = load.get_site_ids(310)
pop_recs = col.defaultdict(dict)
for (site_name, cells), modelname in itt.product(sites.items(), all_models):
print('#####\nreconstituting site {} with model {}\n '.format(
site_name, modelname))
recons_args = {'batch': 310, 'cellid_list': cells, 'modelname': modelname}
recons_cache = ccache.make_cache(
crec.reconsitute_rec,
func_args=recons_args,
classobj_name='reconstitution',
recache=False,
cache_folder='/home/mateo/mycache/reconstitute_recs',
use_hash=True)
reconstituted_recording = ccache.get_cache(recons_cache)
# rasterizes and takes the PSTH of full length signals (containing multiple contex probes)
reconstituted_recording.signals = {
key: val.rasterize()
for key, val in reconstituted_recording.signals.items()
}
reconstituted_recording = npre.average_away_epoch_occurrences(
reconstituted_recording)
pop_recs[site_name][modelname] = reconstituted_recording
# removese the flat response of BRT057b todo solve the bug
del (pop_recs['BRT057b'])
# reorders in dictionary of signals, including only the response and the prediction of each mode
# reformats the epochs
'probe_names': [probe],
'context_transitions': trans_pair,
'probe_order': [probe],
'trans_order': meta['transitions'],
'shuffle_num': meta['nonparam_shuffle'],
'trial_combinations': True
}
shuffled_dispersion_time = make_cache(
function=ndisp.transition_pair_comparison_by_trials,
func_args=func_args,
classobj_name=signal_name,
recache=False,
cache_folder=
f'/home/mateo/mycache/{analysis_name}/{analysis_parameters}')
real, shuffled, scrambled = get_cache(shuffled_dispersion_time)
#define subplots
PairAx = plt.subplot2grid((4, 7), (pp, tt + 1),
rowspan=1,
colspan=1,
fig=fig)
if meta['include_ctx'] is True:
fig, ax = plot.plot_dist_with_CI(real, [shuffled],
['p < 0.05'], ['gray'],
smp_start=0,
smp_end=200,
smp_line=100,
fs=100,
ax=PairAx)
start = 0
end = 200
t1 = (start / 100) - 1
t2 = (end / 100) - 1
##
signal_name = '190431_{}_{}'.format(site, modelname)
func_args = {'signal': sig, 'probe_names': prbs, 'context_names': ctxs, 'shuffle_num': 1000,
'trial_combinations': False}
shuffled_dispersion_time = cch.make_cache(function=cdisp.signal_single_trial_dispersion_pooled_shuffled,
func_args=func_args,
classobj_name=signal_name, recache=False,
cache_folder='/home/mateo/mycache/shuffled_euclideans')
real, distribution = cch.get_cache(shuffled_dispersion_time)
fig, ax = plt.subplots()
line = real[start:end]
shade = distribution[:, start:end]
shade = np.mean(shade, axis=0) + np.std(shade, axis=0) * 2
t = np.linspace(t1, t2, len(line))
ax.plot(t, line , label='{}'.format(site), color='C0')
ax.fill_between(t, -shade , shade , alpha=0.5, color='C0')
ax.axvline(0, color='black', linestyle='--')
# ax.legend(prop={'size': 15})
ax.set_xlabel('time (s)', fontsize=18)
ax.set_ylabel('euclidean distance', fontsize=18)
ax.tick_params(axis='both', which='major', labelsize=15)
func_args = {
'signal': sig,
'probe_names': prbs,
'context_transitions': int_tran,
'channels': what_cells,
'shuffle_num': 1000,
'trial_combinations': True
}
shuffled_dispersion_time = cch.make_cache(
function=cdisp.signal_single_trial_dispersion_pooled_shuffled,
func_args=func_args,
classobj_name=signal_name,
recache=False,
cache_folder='/home/mateo/mycache/shuffled_euclideans')
real, shuffled = cch.get_cache(shuffled_dispersion_time)
########################################################################################################################
# plots only a subset
site = 'AMT029a'
for int_tran in interesting_transition_pairs:
options = {
'batch': 316,
'siteid': site,
'stimfmt': 'envelope',
'rasterfs': 100,
'recache': False,
'runclass': 'CPN',
'stim': False
voc_cmpat = {'voc_0': 'Blues', 'voc_1': 'Oranges', 'voc_2': 'Greens', 'voc_3': 'Reds', 'voc_4': 'Purples'}
sites = load.get_site_ids(310)
################################################
# get and reconstitute single cell recordings into population recording
pop_recs = coll.defaultdict(dict)
for (site_name, cells), modelname in itt.product(sites.items(), all_models):
print('#####\nreconstituting site {} with model {}\n '.format(site_name, modelname))
recons_args = {'batch':310, 'cellid_list':cells, 'modelname': modelname}
recons_cache = ccache.make_cache(crec.reconsitute_rec, func_args=recons_args, classobj_name='reconstitution',
recache=False, cache_folder='/home/mateo/mycache/reconstitute_recs',
use_hash=True)
reconstituted_recording = ccache.get_cache(recons_cache)
pop_recs[site_name][modelname] = reconstituted_recording
# if pop_recs['BRT057b']:
# del(pop_recs['BRT057b'])
################################################
# reorders in dictionary of signals, including only the response and the prediction of each mode
# reformats the epochs
pop_sigs = coll.defaultdict(dict)
for site_key, model_recs in pop_recs.items():
for modelname, rec in model_recs.items():
cpp_rec = cep.set_recording_subepochs(rec, set_pairs=True)
pop_sigs[site_key][modelname] = cpp_rec['pred'].rasterize()
pop_sigs[site_key]['resp'] = cpp_rec['resp'].rasterize()
'probe_names': [probe],
'context_transitions': trans_pair,
'probe_order': [probe],
'trans_order': meta['transitions'],
'shuffle_num': meta['nonparam_shuffle'],
'trial_combinations': True
}
shuffled_dispersion_time = make_cache(
function=ndisp.transition_pair_comparison_by_trials,
func_args=func_args,
classobj_name=signal_name,
recache=False,
cache_folder=
f'/home/mateo/mycache/{analysis_name}/{analysis_parameters}')
real, cont_shuff, pop_shuff = get_cache(shuffled_dispersion_time)
#define subplots
PairAx = plt.subplot2grid((4, 7), (pp, tt + 1),
rowspan=1,
colspan=1,
fig=fig)
fig, ax = plot.plot_dist_with_CI(real, [cont_shuff], ['p < 0.05'],
['gray'],
smp_start=0,
smp_end=100,
smp_line=0,
fs=100,
ax=PairAx)
['{}-{}'.format(key, str(val)) for key, val in meta.items()])
analysis_name = 'CPN_singel_cell_dprime'
cache_folder = pl.Path('C:\\', 'users', 'mateo', 'mycache',
analysis_name, analysis_parameters)
SC_cache = make_cache(function=cell_dprime,
func_args={
'site': site,
'probe': probe,
'meta': meta
},
classobj_name=object_name,
cache_folder=cache_folder,
recache=dprime_recache)
dprime, shuf_dprime, cell_names, trans_pairs = get_cache(SC_cache)
this_site_reals.append(dprime)
this_site_shuffled.append(shuf_dprime)
# single tailed p value base on the montecarlo shuffling
SC_pvalues = np.sum(
(shuf_dprime >= dprime), axis=0) / meta['montecarlo']
this_site_pvalues.append(SC_pvalues)
this_site_reals = np.stack(this_site_reals, axis=0)
this_site_shuffled = np.stack(this_site_shuffled, axis=0)
this_site_pvalues = np.stack(this_site_pvalues, axis=0)
# reorders date in dictionary of cells
for cc, cell in enumerate(cell_names):
best_site = 'BRT056b'
best_model = 'wc.2x2.c-stp.2-fir.2x15-lvl.1-stategain.S-dexp.1'
cells = sites[best_site]
# reconstitutes population recording
print('#####\nreconstituting site {} with model {}\n '.format(
best_site, best_model))
recons_args = {'batch': 310, 'cellid_list': cells, 'modelname': best_model}
recons_cache = ccache.make_cache(
crec.reconsitute_rec,
func_args=recons_args,
classobj_name='reconstitution',
recache=False,
cache_folder='/home/mateo/mycache/reconstitute_recs',
use_hash=True)
rrec = ccache.get_cache(recons_cache)
# formats with CPP epochs and gets
rec = cep.set_recording_subepochs(rrec, set_pairs=False)
# get the envelopes
all_stims = rec['stim'].rasterize().extract_epochs(
['voc_1', 'voc_2', 'voc_3', 'voc_4'])
# gets an example of each stimulus for both channels
all_stims = {stim_name: arr[0, :, :] for stim_name, arr in all_stims.items()}
envelopes = {
'env_0': all_stims['voc_1'][0, :],
'env_1': all_stims['voc_3'][0, :]
}
