def test_fitCumNorm():
#the data are actually from a cum norm so this should be exact
fit = data.FitCumNormal(contrasts, responses, display=0, expectedMin=0.5)
assert numpy.allclose([thresh,sd], fit.params)
modResps = fit.eval(contrasts)
#check that inverse works too
invs = fit.inverse(responses)
assert numpy.allclose(contrasts,invs)#inverse should match the forwards function
#plot if needed
if PLOTTING:
plotFit(modResps, thresh, 'CumulativeNormal (thresh=%.2f, params=%s)' %(fit.inverse(0.75), fit.params))
def test_fitCumNorm():
#the data are actually from a cum norm so this should be exact
fit = data.FitCumNormal(contrasts, responses, display=0, expectedMin=0.5)
assert numpy.allclose([thresh, sd], fit.params)
#check that inverse works too
invs = fit.inverse(responses)
assert numpy.allclose(contrasts,
invs) #inverse should match the forwards function
#plot if needed
if PLOTTING:
pylab.figure(1)
pylab.plot(contrasts, responses, 'o')
pylab.plot(contrasts, fit.eval(contrasts))
pylab.plot([0, thresh], [0.75, 0.75], '--b') #horiz
pylab.plot([thresh, thresh], [0., 0.75], '--b') #vert
pylab.title('Fitting CumulativeNormal')
from psychopy import data
x = [0.1, 0.2, 0.3, 0.4, 0.5, 0.6]
y = [0.1, 0.2, 0.5, 0.7, 0.9, 0.8]
myFunc = data.FitCumNormal(xx=x, yy=y, expectedMin=0)
#myFunc2 = data.FitFunction('logistYN',xx=x,yy=y)
print myFunc.eval(x)
print myFunc.params
print myFunc.inverse(myFunc.eval(x))
"""
[ 0.09031101 0.26949287 0.46916508 0.64717069 0.78403421 0.87766973]
[ 0.1 0.2 0.3 0.4 0.5 0.6]"""
def fitpf(self):
from psychopy import data
import pylab
import matplotlib.pyplot as plt
dfs, _ = self.sumxrl()
allIntensities = dfs.groupby(level=0)['all_intensities'].apply(
np.hstack)
allResponses = dfs.groupby(level=0)['all_responses'].apply(np.hstack)
ntrial = dfs.groupby(level=0)['ntrial'].sum().unique()
num = int(len(allIntensities) / 2)
colorcodes = np.repeat(color4plot(num), 2, axis=0)
# curve fitting and plotting for each condition
fig, axes = plt.subplots(4, 4, figsize=(16, 10))
fig.suptitle(self.sub)
thresh = []
for idx, label in enumerate(allResponses.index):
intensities = abs(allIntensities[label])
responses = allResponses[label]
# bin data and fit to PF
combinedInten, combinedResp, combinedN = data.functionFromStaircase(
intensities, responses,
bins='unique') # optimal: bins='unique'
fit = data.FitCumNormal(combinedInten,
combinedResp,
guess=None,
expectedMin=0.5) # cumulative Gaussian
# fit = data.FitLogistic(combinedInten, combinedResp, guess=None) # logistic
# fit = data.FitWeibull(combinedInten, combinedResp, guess=[3, 0.5], display=1, expectedMin=0.5) #< --- somehow errors
thresh.append(fit.inverse(0.76)) # threshold
print(label + ': % 0.3f' % fit.ssq)
# plot
ax = axes.flatten()[idx]
fontsize = 8
ax.plot(combinedInten,
combinedResp,
'o',
fillstyle='none',
color='grey',
alpha=.5,
markersize=5) # plot combined data points
# ax.plot(intensities, responses, 'o', color='grey', alpha=.5, markersize=3) # plot all data points
# smoothInt = pylab.arange(min(combinedInten), max(combinedInten), 0.1) # x for fitted curve
# smoothResp = fit.eval(smoothInt) # y for fitted curve
smoothResp = pylab.arange(0.5, 1, 0.02)
smoothInt = fit.inverse(smoothResp)
ax.plot(smoothInt, smoothResp, '-',
color=colorcodes[idx]) # plot fitted curve
ax.plot([thresh[idx], thresh[idx]], [0, 0.76], '--', color='grey')
ax.plot([0, thresh[idx]], [0.76, 0.76], '--', color='grey')
ax.set_title(label + ' ' + 'threshold = %0.3f' % thresh[idx],
fontsize=fontsize)
ax.set_ylim([0.5, 1])
ax.set_xlim([0, 6])
ax.tick_params(axis='both', which='major', labelsize=fontsize - 2)
plt.setp(axes[-1, :], xlabel='hue angle')
plt.setp(axes[:, 0], ylabel='correctness')
plt.savefig('pilot_data_fig/' + self.sub + '_' +
self.xstr(self.sel_par) + '_' + self.xstr(self.sel_ses) +
'_' + str(ntrial) + 'trl' + '.pdf')
plt.show()
return allResponses, thresh
def fitData(stairs, percent):
allIntensities, allResponses = [], []
for s in stairs.staircases:
allIntensities.append(s.intensities)
allResponses.append(s.data)
for s in stairs.staircases:
print "Mean for condition ", s.condition['label'], "=", average(
s.reversalIntensities), "std=", std(s.reversalIntensities)
#plot each condition
pylab.subplot(221)
#for stairNumber, thisStair in enumerate(allIntensities):
pylab.plot(allIntensities[0],
'o-',
label=stairs.staircases[0].condition['label'])
pylab.xlabel('Trials')
pylab.ylabel('Speed [cm/s]')
pylab.legend()
# Get combined data
combinedInten, combinedResp, combinedN = data.functionFromStaircase(
allIntensities, allResponses, 10)
# Fit curve - in this case using a Weibull function
fit = data.FitFunction('weibullTAFC',
combinedInten,
combinedResp,
guess=None)
#fit = data.FitCumNormal(combinedInten,combinedResp)
intensitiesDomainInterp = pylab.arange(min(allIntensities[0]),
max(allIntensities[0]), 0.01)
smoothResponses = fit.eval(intensitiesDomainInterp)
thresh = fit.inverse(percent)
#Plot fitted curve
pylab.subplot(222)
pylab.axis(xmin=min(allIntensities[0]), xmax=max(allIntensities[0]))
pylab.xlabel('Speed [cm/s]')
pylab.ylabel('Probability')
pylab.plot(intensitiesDomainInterp, smoothResponses, '-')
pylab.plot([thresh, thresh], [0, percent], '--')
pylab.plot([0, thresh], [percent, percent], '--')
pylab.title('Threshold at ' + str(percent) + '= %0.3f' % (thresh))
# Plot points
pylab.plot(combinedInten, combinedResp, 'o')
pylab.ylim([0, 1])
# SECOND CONDITION, the plots are in a second row, under
pylab.subplot(223)
#for stairNumber, thisStair in enumerate(allIntensities):
pylab.plot(allIntensities[1],
'o-',
label=stairs.staircases[1].condition['label'])
pylab.xlabel('Trials')
pylab.ylabel('Speed [cm/s]')
pylab.legend()
# Get combined data
combinedInten, combinedResp, combinedN = data.functionFromStaircase(
allIntensities[1], allResponses[1], 10)
#fit curve - in this case using a Weibull function
#fit = data.FitFunction('weibullTAFC',combinedInten, combinedResp, guess=None)
fit = data.FitCumNormal(combinedInten, combinedResp)
intensitiesDomainInterp = pylab.arange(min(allIntensities[1]),
max(allIntensities[1]), 0.01)
smoothResponses = fit.eval(intensitiesDomainInterp)
thresh = fit.inverse(percent)
#print "Threshold at " + str(percent) +"% with Cumulative Normal= ",thresh
#Plot fitted curve
pylab.subplot(224)
pylab.axis(xmin=min(allIntensities[1]), xmax=max(allIntensities[1]))
pylab.xlabel('Speed [cm/s]')
pylab.ylabel('Probability')
pylab.plot(intensitiesDomainInterp, smoothResponses, '-')
pylab.plot([thresh, thresh], [0, percent], '--')
pylab.plot([0, thresh], [percent, percent], '--')
pylab.title('Threshold at ' + str(percent) + '= %0.3f' % (thresh))
# Plot points
pylab.plot(combinedInten, combinedResp, 'o')
pylab.ylim([0, 1])
pylab.show()