def processGLMStats(self, trial_FR, trials, welltrain_window=[],correct_resp=[]):
### TODO: when variables are empty
FR_scale=np.concatenate((trial_FR[:,(trials[:,5]==3) & welltrain_window & correct_resp ,:][self.row_sel_3,:,:],
trial_FR[:,(trials[:,5]==6) & welltrain_window & correct_resp,:][self.row_sel_6,:,:]),axis=1)
trial_perf_sel=np.concatenate((trials[(trials[:,5]==3) & welltrain_window & correct_resp,:],
trials[(trials[:,5]==6) & welltrain_window & correct_resp,:]),axis=0)
# S1=4 S2=8 T1=8 T2 =4
trial_S1T1 = (trial_perf_sel[:, 2] == 4) & (trial_perf_sel[:, 3] == 8)
trial_S1T2 = (trial_perf_sel[:, 2] == 4) & (trial_perf_sel[:, 3] == 4)
trial_S2T1 = (trial_perf_sel[:, 2] == 8) & (trial_perf_sel[:, 3] == 8)
trial_S2T2 = (trial_perf_sel[:, 2] == 8) & (trial_perf_sel[:, 3] == 4)
self.S1T1=np.zeros((56,trial_FR.shape[2]))
self.S1T2=np.zeros((56,trial_FR.shape[2]))
self.S2T1=np.zeros((56,trial_FR.shape[2]))
self.S2T2=np.zeros((56,trial_FR.shape[2]))
for su_idx in range(trial_FR.shape[2]):
onesu = np.squeeze(FR_scale[:, :, su_idx]).T
(base_mean, base_std) = baselineVector(onesu)
onesu=(onesu-base_mean)/base_std
self.S1T1[:,su_idx] = np.mean(onesu[trial_S1T1,:],axis=0)
self.S1T2[:,su_idx] = np.mean(onesu[trial_S1T2,:],axis=0)
self.S2T1[:,su_idx] = np.mean(onesu[trial_S2T1,:],axis=0)
self.S2T2[:,su_idx] = np.mean(onesu[trial_S2T2,:],axis=0)
def addPairSelect(self, trial_FR, trial_sel):
trial_sel_left_3 = np.logical_and(
trial_sel[:, 2] != trial_sel[:, 3], trial_sel[:, 5] == 3
)
trial_sel_right_3 = np.logical_and(
trial_sel[:, 2] == trial_sel[:, 3], trial_sel[:, 5] == 3
)
trial_sel_left_6 = np.logical_and(
trial_sel[:, 2] != trial_sel[:, 3], trial_sel[:, 5] == 6
)
trial_sel_right_6 = np.logical_and(
trial_sel[:, 2] == trial_sel[:, 3], trial_sel[:, 5] == 6
)
for su_idx in range(trial_FR.shape[2]):
onesu = np.squeeze(trial_FR[:, :, su_idx]).T
(base_mean, base_std) = baselineVector(onesu)
left_trials_3 = onesu[
trial_sel_left_3, 20:52
] # 28 is test start, aka, -2s to +6s
right_trials_3 = onesu[trial_sel_right_3, 20:52]
left_trials_6 = onesu[trial_sel_left_6, 32:64]
right_trials_6 = onesu[trial_sel_right_6, 32:64]
left_trials = np.concatenate((left_trials_3, left_trials_6))
right_trials = np.concatenate((right_trials_3, right_trials_6))
self.pair_heatmapL.append(
np.mean((left_trials - base_mean) / base_std, axis=0)
)
self.pair_heatmapR.append(
np.mean((right_trials - base_mean) / base_std, axis=0)
)
self.pair_selectivity.append(
self.gauss_average(
(np.mean(left_trials, axis=0) - np.mean(right_trials, axis=0))
/ (
np.mean(right_trials, axis=0)
+ np.mean(left_trials, axis=0)
+ 0.25
)
)
)
# pair stastic selective
bins = np.ones(left_trials.shape[1])
for bin_idx in range(left_trials.shape[1]):
try:
(stat, p) = stats.mannwhitneyu(
left_trials[:, bin_idx].flatten(),
right_trials[:, bin_idx].flatten(),
alternative="two-sided",
)
bins[bin_idx] = p < 0.05
except ValueError:
bins[bin_idx] = 0
self.statistical_pair_selective.append(bins)
def addSampleSelect(self, trial_FR, trial_sel):
trial_sel_left = trial_sel[:, 2] == 4
trial_sel_right = trial_sel[:, 2] == 8
trial_sel_test_left = trial_sel[:, 3] == 4
trial_sel_test_right = trial_sel[:, 3] == 8
trial_sel_3 = trial_sel[:, 5] == 3
trial_sel_6 = trial_sel[:, 5] == 6
for su_idx in range(trial_FR.shape[2]):
onesu = np.squeeze(trial_FR[:, :, su_idx]).T
(base_mean, base_std) = baselineVector(onesu)
left_3_trials = onesu[np.logical_and(trial_sel_left, trial_sel_3), :][:, self.row_sel_3]
left_6_trials = onesu[np.logical_and(trial_sel_left, trial_sel_6), :][:, self.row_sel_6]
right_3_trials = onesu[np.logical_and(trial_sel_right, trial_sel_3), :][:, self.row_sel_3]
right_6_trials = onesu[np.logical_and(trial_sel_right, trial_sel_6), :][:, self.row_sel_6]
left_trials = np.concatenate((left_3_trials, left_6_trials))
right_trials = np.concatenate((right_3_trials, right_6_trials))
left_3_test_trials = onesu[np.logical_and(trial_sel_test_left, trial_sel_3), :][:, self.row_sel_3]
left_6_test_trials = onesu[np.logical_and(trial_sel_test_left, trial_sel_6), :][:, self.row_sel_6]
right_3_test_trials = onesu[np.logical_and(trial_sel_test_right, trial_sel_3), :][:, self.row_sel_3]
right_6_test_trials = onesu[np.logical_and(trial_sel_test_right, trial_sel_6), :][:, self.row_sel_6]
left_test_trials = np.concatenate((left_3_test_trials, left_6_test_trials))
right_test_trials = np.concatenate((right_3_test_trials, right_6_test_trials))
delay_3s_trials = onesu[trial_sel_3, :][:, self.row_sel_3]
delay_6s_trials = onesu[trial_sel_6, :][:, self.row_sel_6]
scaled_FR = np.concatenate((delay_3s_trials, delay_6s_trials))
self.modulation.append(
np.abs(np.mean((scaled_FR - base_mean) / base_std, axis=0))
)
self.heatmapL.append(np.mean((left_trials - base_mean) / base_std, axis=0))
self.heatmapR.append(np.mean((right_trials - base_mean) / base_std, axis=0))
self.test_heatmapL.append(np.mean((left_test_trials - base_mean) / base_std, axis=0))
self.test_heatmapR.append(np.mean((right_test_trials - base_mean) / base_std, axis=0))
self.selectivity.append(
self.gauss_average(
(np.mean(right_trials, axis=0) - np.mean(left_trials, axis=0))
/ (
np.mean(right_trials, axis=0)
+ np.mean(left_trials, axis=0)
+ 0.25
)
)
)
def get_normalized_fr(self, trial_FR, trials, welltrain_window=None, correctResp=None):
### TODO: get normalized FR for epys scaling
for su_idx in range(trial_FR.shape[2]):
onesu = np.squeeze(trial_FR[:, (welltrain_window
& correctResp
& (trials[:, 5] == 6)), su_idx]).T
(base_mean, base_std) = baselineVector(onesu)
allbins = np.mean((onesu - base_mean) / base_std, axis=0)
self.modulation.append(
np.abs([np.mean(allbins[i:i + 4]) for i in np.arange(4, 56, 4)])
)
return np.array(self.modulation).T
def processCTDStats(self, trial_FR, trials, welltrain_window=None, correct_resp=None):
shifted_trial = np.vstack((np.zeros(6), trials[:-1, :]))
### TODO: when variables are empty
FR_D3_S1_S1 = trial_FR[:, np.all(np.vstack(
(trials[:, 5] == 3, trials[:, 2] == 4, shifted_trial[:, 2] == 4, welltrain_window, correct_resp,)),
axis=0, ), :, ]
FR_D3_S2_S1 = trial_FR[
:, np.all(np.vstack(
(trials[:, 5] == 3, trials[:, 2] == 8, shifted_trial[:, 2] == 4, welltrain_window, correct_resp,)),
axis=0, ), :,
]
FR_D3_S1_S2 = trial_FR[
:, np.all(np.vstack(
(trials[:, 5] == 3, trials[:, 2] == 4, shifted_trial[:, 2] == 8, welltrain_window, correct_resp,)),
axis=0, ), :,
]
FR_D3_S2_S2 = trial_FR[
:, np.all(np.vstack(
(trials[:, 5] == 3, trials[:, 2] == 8, shifted_trial[:, 2] == 8, welltrain_window, correct_resp,)),
axis=0, ), :,
]
FR_D6_S1_S1 = trial_FR[
:, np.all(np.vstack(
(trials[:, 5] == 6, trials[:, 2] == 4, shifted_trial[:, 2] == 4, welltrain_window, correct_resp,)),
axis=0, ), :,
]
FR_D6_S2_S1 = trial_FR[
:, np.all(np.vstack(
(trials[:, 5] == 6, trials[:, 2] == 8, shifted_trial[:, 2] == 4, welltrain_window, correct_resp,)),
axis=0, ), :,
]
FR_D6_S1_S2 = trial_FR[
:, np.all(np.vstack(
(trials[:, 5] == 6, trials[:, 2] == 4, shifted_trial[:, 2] == 8, welltrain_window, correct_resp,)),
axis=0, ), :,
]
FR_D6_S2_S2 = trial_FR[
:, np.all(np.vstack(
(trials[:, 5] == 6, trials[:, 2] == 8, shifted_trial[:, 2] == 8, welltrain_window, correct_resp,)),
axis=0, ), :,
]
for su_idx in range(trial_FR.shape[2]):
(mm, std) = baselineVector(
np.squeeze(trial_FR[:, np.logical_and(welltrain_window, correct_resp), su_idx]))
if std > 0 and np.all([x.shape[1] >= 2 for x in (
FR_D3_S1_S1,
FR_D3_S1_S2,
FR_D3_S2_S1,
FR_D3_S2_S2,
FR_D6_S1_S1,
FR_D6_S1_S2,
FR_D6_S2_S1,
FR_D6_S2_S2,
)]):
self.su_list.append(
{
"S1_S1_3m": self.splitMean(FR_D3_S1_S1, su_idx, mm, std),
"S1_S2_3m": self.splitMean(FR_D3_S1_S2, su_idx, mm, std),
"S2_S1_3m": self.splitMean(FR_D3_S2_S1, su_idx, mm, std),
"S2_S2_3m": self.splitMean(FR_D3_S2_S2, su_idx, mm, std),
"S1_S1_6m": self.splitMean(FR_D6_S1_S1, su_idx, mm, std),
"S1_S2_6m": self.splitMean(FR_D6_S1_S2, su_idx, mm, std),
"S2_S1_6m": self.splitMean(FR_D6_S2_S1, su_idx, mm, std),
"S2_S2_6m": self.splitMean(FR_D6_S2_S2, su_idx, mm, std),
}
)
self.avail.append(True)
else:
self.avail.append(False)