proberight = np.where(pside == -1, 1, 0)
pleft_neut = np.where(np.logical_and(pside == 1, cues == 0), 1, 0)
pleft_cued = np.where(np.logical_and(pside == 1, cues == 1), 1, 0)
pright_neut = np.where(np.logical_and(pside == -1, cues == 0), 1, 0)
pright_cued = np.where(np.logical_and(pside == -1, cues == 1), 1, 0)
regressors.append(glm.regressors.CategoricalRegressor(category_list = pleft_neut, codes = 1, name = 'probe left neutral'))
regressors.append(glm.regressors.CategoricalRegressor(category_list = pleft_cued, codes = 1, name = 'probe left cued'))
regressors.append(glm.regressors.CategoricalRegressor(category_list = pright_neut, codes = 1, name = 'probe right neutral'))
regressors.append(glm.regressors.CategoricalRegressor(category_list = pright_cued, codes = 1, name = 'probe right cued'))
error = tfr.metadata.absrdif.to_numpy()
err_npleft = nanzscore(np.where(pleft_neut == 1, error, np.nan))
err_cleft = nanzscore(np.where(pleft_cued == 1, error, np.nan))
err_npright = nanzscore(np.where(pright_neut == 1, error, np.nan))
err_cright = nanzscore(np.where(pright_cued == 1, error, np.nan))
regressors.append(glm.regressors.ParametricRegressor(name = 'err_pleft_neut', values = err_npleft, preproc = None, num_observations = nobs))
regressors.append(glm.regressors.ParametricRegressor(name = 'err_pleft_cued', values = err_cleft, preproc = None, num_observations = nobs))
regressors.append(glm.regressors.ParametricRegressor(name = 'err_pright_neut', values = err_npright, preproc = None, num_observations = nobs))
regressors.append(glm.regressors.ParametricRegressor(name = 'err_pright_cued', values = err_cright, preproc = None, num_observations = nobs))
contrasts = list()
contrasts.append(glm.design.Contrast([ 1, 0, 0, 0, 0, 0, 0, 0], 'pleft_neutral') )
contrasts.append(glm.design.Contrast([ 0, 1, 0, 0, 0, 0, 0, 0], 'pleft_cued') )
contrasts.append(glm.design.Contrast([ 0, 0, 1, 0, 0, 0, 0, 0], 'pright_neutral') )
contrasts.append(glm.design.Contrast([ 0, 0, 0, 1, 0, 0, 0, 0], 'pright_cued') )
pleft_neut = np.where(np.logical_and(pside == 1, cues == 0), 1, 0)
pleft_cued = np.where(np.logical_and(pside == 1, cues == 1), 1, 0)
pright_neut = np.where(np.logical_and(pside == -1, cues == 0), 1, 0)
pright_cued = np.where(np.logical_and(pside == -1, cues == 1), 1, 0)
regressors.append(glm.regressors.CategoricalRegressor(category_list = pleft_neut, codes = 1, name = 'probe left neutral'))
regressors.append(glm.regressors.CategoricalRegressor(category_list = pleft_cued, codes = 1, name = 'probe left cued'))
regressors.append(glm.regressors.CategoricalRegressor(category_list = pright_neut, codes = 1, name = 'probe right neutral'))
regressors.append(glm.regressors.CategoricalRegressor(category_list = pright_cued, codes = 1, name = 'probe right cued'))
prevtrlerr = tfr.metadata.prevtrlerr.to_numpy()
pterr_npleft = nanzscore(np.where(pleft_neut == 1, prevtrlerr, np.nan))
pterr_cleft = nanzscore(np.where(pleft_cued == 1, prevtrlerr, np.nan))
pterr_npright = nanzscore(np.where(pright_neut == 1, prevtrlerr, np.nan))
pterr_cright = nanzscore(np.where(pright_cued == 1, prevtrlerr, np.nan))
regressors.append(glm.regressors.ParametricRegressor(name = 'pterr_pleft_neut', values = pterr_npleft, preproc = None, num_observations = nobs))
regressors.append(glm.regressors.ParametricRegressor(name = 'pterr_pleft_cued', values = pterr_cleft, preproc = None, num_observations = nobs))
regressors.append(glm.regressors.ParametricRegressor(name = 'pterr_pright_neut', values = pterr_npright, preproc = None, num_observations = nobs))
regressors.append(glm.regressors.ParametricRegressor(name = 'pterr_pright_cued', values = pterr_cright, preproc = None, num_observations = nobs))
contrasts = list()
contrasts.append(glm.design.Contrast([ 1, 0, 0, 0, 0, 0, 0, 0], 'pleft_neutral') )
contrasts.append(glm.design.Contrast([ 0, 1, 0, 0, 0, 0, 0, 0], 'pleft_cued') )
contrasts.append(glm.design.Contrast([ 0, 0, 1, 0, 0, 0, 0, 0], 'pright_neutral') )
glmdata = glm.data.TrialGLMData(data = epochs.get_data(), time_dim = 2, sample_rate = 500)
nobs = glmdata.num_observations
trials = np.ones(nobs)
cues = epochs.metadata.cue.to_numpy()
error = epochs.metadata.absrdif.to_numpy()
confwidth = epochs.metadata.confwidth.to_numpy() #confidence width in degrees, higher = less confident
conf = np.radians(np.multiply(confwidth, -1)) #reverse the sign so higher (less negative) = more confident, then convert to radians so same scale as error
confdiff = np.radians(epochs.metadata.confdiff.to_numpy()) #error awareness (metacognition) on that trial (or prediction error)
neutral = np.where(cues == 0, 1, 0)
targinconf = np.less_equal(confdiff,0)
targoutsideconf = np.greater(confdiff,0)
incorrvscorr = np.where(targinconf == 0, 1, -1)
errorcorr = nanzscore(np.where(targinconf == 1, error, np.nan))
errorincorr = nanzscore(np.where(targoutsideconf == 1, error, np.nan))
confcorr = nanzscore(np.where(targinconf == 1, conf, np.nan))
confincorr = nanzscore(np.where(targoutsideconf == 1, conf, np.nan))
pside = epochs.metadata.pside.to_numpy()
pside = np.where(pside == 0, 1, -1)
#add regressors to the model
regressors = list()
regressors.append(glm.regressors.CategoricalRegressor(category_list = targinconf, codes = 1, name = 'correct'))
regressors.append(glm.regressors.CategoricalRegressor(category_list = targoutsideconf, codes = 1, name = 'incorrect'))
regressors.append( glm.regressors.ParametricRegressor(name = 'error-correct', values = errorcorr, preproc = None, num_observations = nobs))
regressors.append( glm.regressors.ParametricRegressor(name = 'error-incorrect', values = errorincorr, preproc = None, num_observations = nobs))
regressors.append(
glm.regressors.CategoricalRegressor(category_list=pright_neut,
codes=1,
name='probe right neutral'))
regressors.append(
glm.regressors.CategoricalRegressor(category_list=pright_cued,
codes=1,
name='probe right cued'))
cw = tfr.metadata.confwidth.to_numpy()
conf = np.multiply(
cw, -1
) #make this regressor indicate confidence, not confidence width.
#multiply by -1 as large confidence widths = low confidence. now lower numbers = lower confidence
conf_npleft = nanzscore(np.where(pleft_neut == 1, conf, np.nan))
conf_cleft = nanzscore(np.where(pleft_cued == 1, conf, np.nan))
conf_npright = nanzscore(np.where(pright_neut == 1, conf, np.nan))
conf_cright = nanzscore(np.where(pright_cued == 1, conf, np.nan))
regressors.append(
glm.regressors.ParametricRegressor(name='conf_pleft_neut',
values=conf_npleft,
preproc=None,
num_observations=nobs))
regressors.append(
glm.regressors.ParametricRegressor(name='conf_pleft_cued',
values=conf_cleft,
preproc=None,
num_observations=nobs))
regressors.append(
error = epochs.metadata.absrdif.to_numpy()
confwidth = epochs.metadata.confwidth.to_numpy(
) #confidence width in degrees, higher = less confident
conf = np.radians(
np.multiply(confwidth, -1)
) #reverse the sign so higher (less negative) = more confident, then convert to radians so same scale as error
confdiff = np.radians(
epochs.metadata.confdiff.to_numpy()
) #error awareness (metacognition) on that trial (or prediction error)
neutral = np.where(cues == 0, 1, 0)
targinconf = np.less_equal(confdiff, 0)
targoutsideconf = np.greater(confdiff, 0)
incorrvscorr = np.where(targinconf == 0, 1, -1)
errorcorr = nanzscore(np.where(targinconf == 1, error, np.nan))
errorincorr = nanzscore(np.where(targoutsideconf == 1, error, np.nan))
confcorr = nanzscore(np.where(targinconf == 1, conf, np.nan))
confincorr = nanzscore(np.where(targoutsideconf == 1, conf, np.nan))
pside = epochs.metadata.pside.to_numpy()
pside = np.where(pside == 0, 1, -1)
#add regressors to the model
if model == 1:
glm_folder = 'epochs_glm5'
regressors = list()
regressors.append(
glm.regressors.CategoricalRegressor(category_list=targinconf,
codes=1,
glmdata = glm.data.TrialGLMData(data=tfr.data, time_dim=3, sample_rate=100)
nobs = glmdata.num_observations
trials = np.ones(
glmdata.num_observations) #regressor for just grand mean response
error = tfr.metadata.absrdif.to_numpy()
confwidth = tfr.metadata.confwidth.to_numpy(
) #confidence width in degrees, higher = less confident
conf = np.radians(
np.multiply(confwidth, -1)
) #reverse the sign so higher (less negative) = more confident, then convert to radians so same scale as error
confdiff = tfr.metadata.confdiff.to_numpy(
) #error awareness (metacognition) on that trial (or prediction error)
err = nanzscore(error) #dont worry about nans this is just normal zscore
cw = nanzscore(confwidth)
confidence = nanzscore(conf)
regressors = list()
regressors.append(
glm.regressors.ParametricRegressor(name='mean induced',
values=trials,
preproc=None,
num_observations=nobs))
regressors.append(
glm.regressors.ParametricRegressor(name='error',
values=err,
preproc=None,
num_observations=nobs))
regressors.append(
confdiff = epochs.metadata.confdiff.to_numpy(
) #error awareness on that trial
neutral = np.where(cues == 0, 1, 0)
nxttrlcw = epochs.metadata.nexttrlcw.to_numpy()
nxttrlconf = np.multiply(nxttrlcw, -1)
# trlupdate = np.subtract(confwidth, nxttrlcw) #positive value means that the confidence with got narrower -- negative value means it got wider
trlupdate = np.subtract(
nxttrlconf, conf
) #positive values mean they became more confident, negative values they became less confident
targinconf = np.less_equal(confdiff, 0)
targoutsideconf = np.greater(confdiff, 0)
incorrvscorr = np.where(targinconf == 0, 1, -1)
confwidth = nanzscore(confwidth)
conf = nanzscore(conf)
confdiff = nanzscore(confdiff)
trlupdate = nanzscore(trlupdate)
regressors = list()
contrasts = list()
if glmnum == 7:
regressors.append(
glm.regressors.ParametricRegressor(name='trials',
values=trials,
preproc=None,
num_observations=nobs))
regressors.append(
glm.regressors.ParametricRegressor(name='update',
values=trlupdate,
neutral = np.where(cues == 0, 1, 0)
targinconf = np.less_equal(confdiff, 0)
targoutsideconf = np.greater(confdiff, 0)
threshold = -10
if threshold == -6:
glmnum = 5
else:
glmnum = 6
defscorr = np.less_equal(
confdiff, threshold) #easily correct (over 5 degrees underconfident)
justcorr = np.isin(
confdiff, np.arange(threshold + 1, 1)
) #within 5 degrees of the boundary (less than 5 degrees underconfident)
err_defscorr = nanzscore(np.where(defscorr == 1, error, np.nan))
err_justcorr = nanzscore(np.where(justcorr == 1, error, np.nan))
err_incorr = nanzscore(np.where(targoutsideconf == 1, error, np.nan))
conf_defscorr = nanzscore(np.where(defscorr == 1, conf, np.nan))
conf_justcorr = nanzscore(np.where(justcorr == 1, conf, np.nan))
conf_incorr = nanzscore(np.where(targoutsideconf == 1, conf, np.nan))
pside = epochs.metadata.pside.to_numpy()
pside = np.where(pside == 0, 1, -1)
#add regressors to the model
regressors = list()
regressors.append(
glm.regressors.CategoricalRegressor(category_list=defscorr,
regressors.append(
glm.regressors.CategoricalRegressor(category_list=pleft_cued,
codes=1,
name='probe left cued'))
regressors.append(
glm.regressors.CategoricalRegressor(category_list=pright_neut,
codes=1,
name='probe right neutral'))
regressors.append(
glm.regressors.CategoricalRegressor(category_list=pright_cued,
codes=1,
name='probe right cued'))
dt = tfr.metadata.DT.to_numpy()
dt_npleft = nanzscore(np.where(pleft_neut == 1, dt, np.nan))
dt_cleft = nanzscore(np.where(pleft_cued == 1, dt, np.nan))
dt_npright = nanzscore(np.where(pright_neut == 1, dt, np.nan))
dt_cright = nanzscore(np.where(pright_cued == 1, dt, np.nan))
regressors.append(
glm.regressors.ParametricRegressor(name='dt_pleft_neut',
values=dt_npleft,
preproc=None,
num_observations=nobs))
regressors.append(
glm.regressors.ParametricRegressor(name='dt_pleft_cued',
values=dt_cleft,
preproc=None,
num_observations=nobs))
regressors.append(