plt.subplot(1,1,1)
plt.plot(StimLevels, pObs,'*r')
# Plot a smoother fitted function
smoothrang = np.linspace(StimLevels[0], StimLevels[-1], 100 )
LogLikX, smoothprob=ProbitLogit(params0, smoothrang, NumPos, Ntrials, LowerAsymptote, ProbitOrLogit,0)
plt.plot(smoothrang, smoothprob, '--b', label='init conds');
#plt.plot(StimLevels,p0,'.--b', label='initial conditions')
plt.title('Fit data');
plt.xlabel('Stimulus Intensity');
plt.ylabel('Probability Correct')
## now do the search
#[params,chisqLL]=fminsearch('ProbitLogit',params0,[], StimLevels, NumPos, Ntrials,
# LowerAsymptote, ProbitOrLogit,1)
out = fitpf(params0, StimLevels, NumPos, Ntrials, LowerAsymptote,
ProbitOrLogit, output_param_search=True)
pfinal = out[0] # Y
params0 = out[0]
searched_params = np.array( out[1] )
LogLikf, probExpect=ProbitLogit(pfinal, StimLevels, NumPos, Ntrials, LowerAsymptote, ProbitOrLogit,1)
if plot_opt in ('both','pf'):
# Plot a smoother fitted function
smoothrang = np.linspace(StimLevels[0], StimLevels[-1], 100 )
LogLikX, smoothprob=ProbitLogit(pfinal, smoothrang, NumPos, Ntrials, LowerAsymptote, ProbitOrLogit,0)
plt.plot(smoothrang, smoothprob, '-b', label='LL search');
error=np.sqrt(probExpect*(1-probExpect)/Ntrials);
plt.errorbar(StimLevels,probExpect,error, fmt=None, ecolor='b');
plt.figure()
StimLevels=np.arange(6)
Ntrials=np.ones(6)
ProbitOrLogit=2
LowerAsymptote=0.5
plt.plot(StimLevels,hw2res[3],'b*', label='$Pc$')
plt.plot(StimLevels,unb[0],'rx', label='$Pc_{max}$')
smoothrang = np.linspace(StimLevels[0], StimLevels[-1], 100 )
# Fit to Pc
# 2 = normal slope (Palamedes way)
# [3,2.0] = guess
paramsfit = pf.fitpf( [3,2.0], StimLevels, hw2res[3], Ntrials, LowerAsymptote, ProbitOrLogit)
# Plot a smoother fitted function
LogLikX, smoothprob=pf.ProbitLogit(paramsfit, smoothrang, hw2res[3], Ntrials, LowerAsymptote, ProbitOrLogit,0)
plt.plot(smoothrang, smoothprob, 'b--', label='Fit to $Pc$');
print "Fit to Pc: %s" % str(paramsfit)
# Fit to Pc_max
paramsfit2 = pf.fitpf( [3,2.0], StimLevels, unb[0], Ntrials, LowerAsymptote, ProbitOrLogit)
LogLikX, smoothprob=pf.ProbitLogit(paramsfit2, smoothrang, hw2res[3], Ntrials, LowerAsymptote, ProbitOrLogit,0)
plt.plot(smoothrang, smoothprob, 'r--', label='Fit to $Pc_{max}$');
print "Fit to Pc_max: %s" % str(paramsfit2)
plt.legend(loc='best')
plt.show()
#fake_dprime = fake_dprime( prins_tab62_h, prins_tab62_f )
