try:
lapse
except NameError:
lapse = 0.00
doPlots = False
gridGrain = 150
numlevels = 6
trials_per = trials/numlevels
alpha = 10.0
beta = 3.0
levels_frac = np.array([.1,.3,.4,.6,.7,.9]) # Prins' s2
params2=[alpha,beta,0,0] # Wichmann/Prins use this to gen some initial values
params=[alpha,beta,0.5,lapse]
alphas = np.logspace( *np.log10( pf.fn_weibull_inv(np.array([0.01,0.99]), params2) ), num=gridGrain )
betas = np.logspace( -1.0, 2.0, gridGrain)
gamma = 0.5
lambdas = np.arange( 0, .06+np.finfo(float).eps, 0.005)
# Model:
fit1 = pf.pf_generic( pf.fn_weibull, params, [0,1,3])
# Data:
levels = pf.fn_weibull_inv( levels_frac, params2 )
trials_arr = np.tile( trials_per, numlevels)
data=pf.experiment(levels, trials_arr, pcorr=pf.fn_weibull( levels, [alpha,beta,0.5,0]) )
data.simulate()
# Rebuild the huge grid of initial LogPs if we need to
# Takes about 16.5 secs on my Thinkpad x60
def param_scatter(params, true_params, idxOfLapse=3, ax=None, needToConvert=False, plot_marginals=False):
global axMargx, axMargy
if needToConvert:
fit_params_adj = np.array( [pf.convertToWichmann( p ) for p in params] )
else:
fit_params_adj = np.copy(params)
true_pse = pf.fn_weibull_inv(0.5, true_params )
true_slope = pf.fn_weibull_deriv( true_pse, true_params )
laps_lowers = np.where( fit_params_adj[:,idxOfLapse] < 0.0001 )[0]
laps_uppers = np.where( fit_params_adj[:,idxOfLapse] >= 0.0599)[0]
laps_mids = np.where( np.all((fit_params_adj[:,idxOfLapse] >= 0.0001,fit_params_adj[:,idxOfLapse] < 0.0599),0) )[0]
if ax==None:
fig=plt.figure()
ax = fig.add_subplot(1,1,1)
if plot_marginals:
divider = make_axes_locatable( ax)
# create a new axes with a height of 1.2 inch above the axScatter
axMargx = divider.new_vertical(1.2, pad=0.22, sharex=ax)
# create a new axes with a width of 1.2 inch on the right side of the
# axScatter
# TODO: Figure out how to make the hists not overlap with the scatter
axMargy = divider.new_horizontal(1.2, pad=0.22, sharey=ax)
ax.plot( fit_params_adj[laps_lowers,0], fit_params_adj[laps_lowers,1], 'bx', label='$\lambda_{est}=0$' )
ax.plot( fit_params_adj[laps_mids,0], fit_params_adj[laps_mids,1], 'g*', label='$\lambda_{est}\/{o.w.}$' )
ax.plot( fit_params_adj[laps_uppers,0], fit_params_adj[laps_uppers,1], 'r.', label='$\lambda_{est}=0.6$' )
ax.legend( loc='upper left')
ax.axis('tight')
x = [fit_params_adj[:,0].min(), fit_params_adj[:,0].max()]
y = [fit_params_adj[:,1].min(), fit_params_adj[:,1].max()]
ax.plot( [true_pse, true_pse], [fit_params_adj[:,1].min(), fit_params_adj[:,1].max()], 'k:' )
ax.plot( [fit_params_adj[:,0].min(), fit_params_adj[:,0].max()], [true_slope, true_slope], 'k:' )
ax.set_xlabel('pse (~%.4g)' % true_pse )
ax.set_ylabel('slope@pse (~%.4g)' % true_slope)
ax.loglog()
mean0 = np.mean( fit_params_adj[laps_lowers], 0 )
mean1 = np.mean( fit_params_adj[laps_mids], 0 )
mean2 = np.mean( fit_params_adj[laps_uppers], 0 )
xbins = 30
ybins = 30
if plot_marginals:
fig.add_axes(axMargx)
axMargx.hist( fit_params_adj[:,0], bins=xbins )
fig.add_axes(axMargy)
ax.axis('tight')
axMargy.set_yscale("log", nonposy='clip')
bins = 10**np.linspace( np.log10(y[0]), np.log10(y[1]), num=ybins )
axMargy.hist( fit_params_adj[:,1],orientation='horizontal', bins=bins )
oldticks = axMargy.get_xticklabels()
[tick.set_rotation(-90) for tick in oldticks]
ax.axis('tight')
covarp=np.cov(fit_params_adj.T);
sep=np.sqrt(np.diag(covarp))
corp=covarp/(sep.T*sep);
corrp=[corp[1,0],corp[idxOfLapse,0],corp[idxOfLapse,1]]
covar0 = np.cov(fit_params_adj[laps_lowers].T); se0 = np.sqrt( np.diag( covar0))
covar1 = np.cov(fit_params_adj[laps_mids].T); se1 = np.sqrt( np.diag( covar1))
covar2 = np.cov(fit_params_adj[laps_uppers].T); se2 = np.sqrt( np.diag( covar2))
tiny=8
covartop = 0.3
hite = 0.05
ax.text( 0.05, covartop, r'$corr_{\alpha,\beta}$=%.4g' % corrp[0],
horizontalalignment='left', verticalalignment='top', transform=ax.transAxes)
ax.text( 0.05, covartop-hite, r'$corr_{\alpha,\lambda}$=%.4g' % corrp[1],
horizontalalignment='left', verticalalignment='top', transform=ax.transAxes)
ax.text( 0.05, covartop-2*hite, r'$corr_{\beta,\lambda}$=%.4g' % corrp[2],
horizontalalignment='left', verticalalignment='top', transform=ax.transAxes)
parmstop = 0.12
hite2 = 0.04
ax.text( 0.05, parmstop, '$\mu_{0.0}$=%s' % (mean_se_string(mean0, se0)),
horizontalalignment='left', verticalalignment='top', transform=ax.transAxes, size=tiny)
ax.text( 0.05, parmstop-hite2, '$\mu$=%s' % (mean_se_string(mean1, se1)),
horizontalalignment='left', verticalalignment='top', transform=ax.transAxes, size=tiny)
ax.text( 0.05, parmstop-hite2*2, '$\mu_{0.06}$=%s' % (mean_se_string(mean2, se2)),
horizontalalignment='left', verticalalignment='top', transform=ax.transAxes, size=tiny)
return covarp, sep, corp, corrp
