def test_deviance_precalc(self):
rpar = pSSLBA_modelA.paramspec(*list(np.random.rand(8)))
self.mod.set_params_c(rpar)
L1 = self.mod.deviance_precalc(self.ds)
self.mod.set_params(rpar)
L2 = self.mod.deviance(self.ds)
assert abs(L1 - L2) < 1e-10, "L1,L2=%f, %f" % (L1, L2)
def test_loglikelihood(self):
factors=[{'sleepdep':['normal','deprived']},
{'stimulus':['left', 'right']}]
responses=['left','right']
design=Design(factors,responses, 'stimulus')
dat=pd.read_csv('./data/sleep_stop_onesubj_test.csv')
dat.columns=['sleepdep','stimulus','SSD','response','correct', 'RT']
ds=StopTaskDataSet(design,dat)
pars=pSSLBA_modelA.paramspec(.1, .2, .2, .5, .8, 2, 1, 0)
mod=pSSLBA_modelA(design, pars)
print mod
Ls=StopTaskRaceModel.deviance(mod, ds)
L2=mod.deviance(ds)
assert np.abs(Ls-L2)<1e-8
def setUp(self):
self.factors = [{
'sleepdep': ['normal', 'deprived']
}, {
'stimulus': ['left', 'right']
}]
self.responses = ['left', 'right']
self.design = Design(self.factors, self.responses, 'stimulus')
self.dat = pd.read_csv('./data/sleep_stop_onesubj_test.csv')
assert self.dat.shape[0] == 800
self.dat.columns = [
'sleepdep', 'stimulus', 'SSD', 'response', 'correct', 'RT'
]
self.ds = StopTaskDataSet(self.design, self.dat)
pars = pSSLBA_modelA.paramspec(.2, .15, .2, 1.0, 1.0, 2, 1, 0.5)
self.mod = pSSLBA_modelA(self.design, pars)
def test_set_params_c(self):
#rpar=np.random.rand(8)
rpar = pSSLBA_modelA.paramspec(*list(np.random.rand(8)))
self.mod.set_params_c(rpar)
cp1 = self.mod.cpars
self.mod.set_params(rpar)
cp2 = self.mod.get_cpars()
parnames = [
'go_v', 'go_ter', 'go_A', 'go_b', 'go_sv', 'stop_v', 'stop_ter',
'stop_A', 'stop_b', 'stop_sv', 'pgf', 'ptf'
]
ix = 0
for c1, c2 in zip(cp1, cp2):
assert np.all(
np.abs(c1 - c2) < 1e-10), "%i (=parameter %s): %s,%s" % (
ix, parnames[ix], str(c1), str(c2))
ix += 1
def test_sslba_likelihood_trials(self):
factors=[{'sleepdep':['normal','deprived']},
{'stimulus':['left', 'right']}]
responses=['left','right']
design=Design(factors,responses, 'stimulus')
dat=pd.read_csv('./data/sleep_stop_onesubj_test.csv')
dat.columns=['sleepdep','stimulus','SSD','response','correct', 'RT']
ds=StopTaskDataSet(design,dat)
pars=pSSLBA_modelA.paramspec(.1, .2, .2, .5, .8, 2, 1, 0)
mod=pSSLBA_modelA(design, pars)
print mod
Ls=StopTaskRaceModel.likelihood_trials(mod, ds)
L2=mod.likelihood_trials(ds)
goodix=(np.abs(Ls-L2)<1e-5)
badix=np.logical_not(goodix)
assert np.sum(badix)==0, "num bad idx=%i"%np.sum(badix)
def test_loglikelihood(self):
factors = [{
'sleepdep': ['normal', 'deprived']
}, {
'stimulus': ['left', 'right']
}]
responses = ['left', 'right']
design = Design(factors, responses, 'stimulus')
dat = pd.read_csv('./data/sleep_stop_onesubj_test.csv')
dat.columns = [
'sleepdep', 'stimulus', 'SSD', 'response', 'correct', 'RT'
]
ds = StopTaskDataSet(design, dat)
pars = pSSLBA_modelA.paramspec(.1, .2, .2, .5, .8, 2, 1, 0)
mod = pSSLBA_modelA(design, pars)
print mod
Ls = StopTaskRaceModel.deviance(mod, ds)
L2 = mod.deviance(ds)
assert np.abs(Ls - L2) < 1e-8
def test_init(self):
factors = [{
'sleepdep': ['normal', 'deprived']
}, {
'stimulus': ['left', 'right']
}]
responses = ['left', 'right']
design = Design(factors, responses, 'stimulus')
dat = pd.read_csv('./data/sleep_stop_onesubj_test.csv')
assert dat.shape[0] == 800
dat.columns = [
'sleepdep', 'stimulus', 'SSD', 'response', 'correct', 'RT'
]
ds = StopTaskDataSet(design, dat)
pars = pSSLBA_modelA.paramspec(.2, .15, .2, 1.0, 1.0, 2, 1, 0.5)
mod = pSSLBA_modelA(design, pars)
print mod.parstring(full=True)
print mod
print mod.go_accumulators[0]
nsamples = 100
x = np.linspace(0.01, 10, nsamples)
condition = 0
dens = mod.dens_acc_go(x, condition, 1)
cscore = 0
for i in range(design.nresponses()):
score = scipy.integrate.quad(mod.dens_acc_go,
0.01,
np.infty,
args=(condition, i))[0]
print "Chance of winning Acc %i (condition=%s): %f" % (
i, str(design.condidx(condition)), score)
cscore += score
assert abs(cscore - 1) < 1e-6
print cscore
L = mod.loglikelihood(ds)
assert np.isfinite(L)
print "L=", L
def test_sslba_likelihood_trials(self):
factors = [{
'sleepdep': ['normal', 'deprived']
}, {
'stimulus': ['left', 'right']
}]
responses = ['left', 'right']
design = Design(factors, responses, 'stimulus')
dat = pd.read_csv('./data/sleep_stop_onesubj_test.csv')
dat.columns = [
'sleepdep', 'stimulus', 'SSD', 'response', 'correct', 'RT'
]
ds = StopTaskDataSet(design, dat)
pars = pSSLBA_modelA.paramspec(.1, .2, .2, .5, .8, 2, 1, 0)
mod = pSSLBA_modelA(design, pars)
print mod
Ls = StopTaskRaceModel.likelihood_trials(mod, ds)
L2 = mod.likelihood_trials(ds)
goodix = (np.abs(Ls - L2) < 1e-5)
badix = np.logical_not(goodix)
assert np.sum(badix) == 0, "num bad idx=%i" % np.sum(badix)