def predict(params, data, simfx=sims.sim_exp, ntrials=160, pslow=0.0, pfast=0.0, nsims=100, nsims_per_sub=1, errors=False, save=False, RTname='RT_simexp', ACCname='Acc_simexp'):
"""
Arguments:
params (dict): hierarchical dictionary created with
either parse_stats() or reformat_sims_input()
data (pandas df): pandas dataframe with the empirical data used
to fit the model to generate the simulation parameters
*Simulates subject-wise data using parameter estimates for each subj/condition
and calls rt and acc plotting functions to plot behavioral data against model predictions
*If save=True, will save RT and ACC plots to working dir
"""
from myhddm import parse
simdf_list=[]
if len(data.cue.unique())==3:
x=np.array([1,2,3])
code_type='HNL'
else:
x=np.array([1, 2, 3, 4, 5])
code_type='AllP'
if errors:
face_rt, house_rt=parse.get_emp_error_rt(data=data)
face_acc, house_acc, cf, ch=parse.get_empirical_means(data=data, code_type=code_type)
else:
face_acc, house_acc, face_rt, house_rt=parse.get_empirical_means(data=data, code_type=code_type)
face_rt_error, house_rt_error=parse.get_emp_error_rt(data=data)
sem_list=parse.get_emp_SE(data, code_type)
#init sep figure, axes for RT & ACC data
fig_rt, ax_rt=plt.subplots(1)
fig_acc, ax_acc=plt.subplots(1)
fig_acc.subplots_adjust(top=0.9, left=0.15, right=0.88, bottom=0.15)
fig_rt.subplots_adjust(top=0.9, left=0.15, right=0.88, bottom=0.15)
flast_rt=np.zeros([5])
hlast_rt=np.zeros([5])
flast_acc=np.zeros([5])
hlast_acc=np.zeros([5])
for i in range(nsims):
simdf, params_used=simfx(pdict=params, ntrials=ntrials, pfast=pfast, pslow=pslow, nsims_per_sub=nsims_per_sub)
simdf['sim_n']=[i]*len(simdf.index)
simdf_list.append(simdf)
Ftheo_acc, Htheo_acc = parse.get_theo_acc(simdf=simdf, code_type=code_type)
flast_acc, hlast_acc = pred_accPLOT(code_type=code_type, acc_ax=ax_acc, xacc=x, yaccFace=Ftheo_acc, yaccHouse=Htheo_acc, ind=i, flast_acc=flast_acc, hlast_acc=hlast_acc)
if errors:
Ftheo_rt, Htheo_rt=parse.get_theo_error_rt(simdf=simdf)
else:
Ftheo_rt, Htheo_rt=parse.get_theo_rt(simdf=simdf, code_type=code_type)
flast_rt, hlast_rt = pred_rtPLOT(code_type=code_type, rt_ax=ax_rt, xrt=x, yrtFace=Ftheo_rt, yrtHouse=Htheo_rt, ind=i, flast_rt=flast_rt, hlast_rt=hlast_rt)
simdf_concat=pd.concat(simdf_list)
#plot empirical ACC
#ax_acc.grid()
#f_emp_acc=ax_acc.errorbar(x, face_acc, yerr=sem_list[0], elinewidth=3.5, ecolor='k', color='Blue', lw=6.0)
#h_emp_acc=ax_acc.errorbar(x, house_acc, yerr=sem_list[1], elinewidth=3.5, ecolor='k', color='Green', lw=6.0)
f_emp_acc=ax_acc.errorbar(x, face_acc, yerr=sem_list[0], elinewidth=3.5, ecolor='k', color='Blue', lw=6.0)
h_emp_acc=ax_acc.errorbar(x, house_acc, yerr=sem_list[1], elinewidth=3.5, ecolor='k', color='Red', lw=6.0)
#ax_acc.legend((ax_acc.lines[-4], ax_acc.lines[-1], ax_acc.lines[0], ax_acc.lines[1]), ('Face Data', 'House Data', 'Face Model', 'House Model'), loc=0, fontsize=18)
#plot empirical RT
#ax_rt.grid()
#f_emp_rt=ax_rt.errorbar(x, face_rt, yerr=sem_list[2], elinewidth=3.5, ecolor='k', color='Blue', lw=6.0)
#h_emp_rt=ax_rt.errorbar(x, house_rt, yerr=sem_list[3], elinewidth=3.5, ecolor='k', color='Green', lw=6.0)
f_emp_rt=ax_rt.errorbar(x, face_rt, yerr=sem_list[2], elinewidth=3.5, ecolor='k', color='Blue', lw=6.0)
h_emp_rt=ax_rt.errorbar(x, house_rt, yerr=sem_list[3], elinewidth=3.5, ecolor='k', color='Red', lw=6.0)
#ax_rt.legend((ax_rt.lines[-4], ax_rt.lines[-1], ax_rt.lines[0], ax_rt.lines[1]), ('Face Data', 'House Data', 'Face Model', 'House Model'), loc=0, fontsize=18)
simdf_concat.to_csv("simdf.csv")
if save:
fig_rt.savefig(RTname+'.jpeg', dpi=900)
fig_acc.savefig(ACCname+'.jpeg', dpi=900)
#fig_rt.savefig(RTname+'.png', format='png', dpi=500)
#fig_acc.savefig(ACCname+'.png', format='png', dpi=500)
#fig_rt.savefig(RTname+'.tif', format='tif', dpi=500)
#fig_acc.savefig(ACCname+'.tif', fortmat='tif', dpi=500)
return simdf_concat
def plot_data(data, save=True, RTname='RT_Data', ACCname='Acc_Data'):
"""
Arguments:
params (dict): hierarchical dictionary created with
either parse_stats() or reformat_sims_input()
data (pandas df): pandas dataframe with the empirical data used
to fit the model to generate the simulation parameters
*Simulates subject-wise data using parameter estimates for each subj/condition
and calls rt and acc plotting functions to plot behavioral data against model predictions
*If save=True, will save RT and ACC plots to working dir
"""
sns.set_style("white")
#sns.despine()
if len(data.cue.unique())==3:
x=np.array([1,2,3])
code_type='HNL'
else:
x=np.array([1, 2, 3, 4, 5])
code_type='AllP'
face_acc, house_acc, face_rt, house_rt=parse.get_empirical_means(data=data, code_type=code_type)
sem_list=parse.get_emp_SE(data, code_type)
#init sep figure, axes for RT & ACC data
fig_rt, ax_rt=plt.subplots(1)
sns.despine()
fig_acc, ax_acc=plt.subplots(1)
sns.despine()
fig_acc.subplots_adjust(top=0.9, left=0.15, right=0.88, bottom=0.15)
fig_rt.subplots_adjust(top=0.9, left=0.15, right=0.88, bottom=0.15)
#fat
#fig_rt.subplots_adjust(top=0.7, left=0.15, right=0.88, bottom=0.15)
#plot empirical ACC
f_emp_acc=ax_acc.errorbar(x, face_acc, yerr=sem_list[0], elinewidth=2.5, ecolor='k', color='Blue', lw=4.0)
h_emp_acc=ax_acc.errorbar(x, house_acc, yerr=sem_list[1], elinewidth=2.5, ecolor='k', color='Red', lw=4.0)
#ax_acc.set_title("Accuracy")
ax_acc.legend((ax_acc.lines[-4], ax_acc.lines[-1]), ('Face', 'House'), loc=0, fontsize=18)
#plot empirical RT
f_emp_rt=ax_rt.errorbar(x, face_rt, yerr=sem_list[2], elinewidth=2.5, ecolor='k', color='Blue', lw=4.0)
h_emp_rt=ax_rt.errorbar(x, house_rt, yerr=sem_list[3], elinewidth=2.5, ecolor='k', color='Red', lw=4.0)
#ax_rt.set_title("Response-Time")
ax_rt.legend((ax_rt.lines[-4], ax_rt.lines[-1]), ('Face', 'House'), loc=0, fontsize=18)
if code_type=='HNL':
ax_rt.set_ylim(1.5, 3.5)
ax_rt.set_xticks([1, 2, 3])
ax_rt.set_xlim(0.5, 3.5)
ax_rt.set_xticklabels(['80H', '50N', '80F'], fontsize=32)
ax_rt.set_ylabel('Response Time (s)', fontsize=35, labelpad=14)
ax_rt.set_xlabel('Prior Probability Cue', fontsize=35, labelpad=10)
ax_rt.set_yticklabels(np.arange(1.5, 4, 0.5), fontsize=25)
ax_acc.set_ylim(0.7, 1.0)
ax_acc.set_xticks([1, 2, 3])
ax_acc.set_xlim(0.5, 3.5)
ax_acc.set_xticklabels(['80H', '50N', '80F'], fontsize=32)
ax_acc.set_ylabel('Proportion Correct', fontsize=35, labelpad=14)
ax_acc.set_xlabel('Prior Probability Cue', fontsize=35, labelpad=10)
ax_acc.set_yticks(np.arange(0.6, 1.05, .05))
ax_acc.set_yticklabels(np.arange(0.6, 1.05, .05), fontsize=25)
else:
ax_rt.set_ylim(1.6, 3.5)
ax_rt.set_xticks([1, 2, 3, 4, 5])
ax_rt.set_xlim(0.6, 5.5)
ax_rt.set_xticklabels(['90H', '70H', '50/50', '70F', '90F'], fontsize=18)
ax_rt.set_yticks(np.arange(1.5, 4.0, 0.5))
ax_rt.set_yticklabels(np.arange(1.5, 4.0, 0.5), fontsize=18)
ax_rt.set_ylabel('Response Time (s)', fontsize=22, labelpad=14)
ax_rt.set_xlabel('Prior Probability Cue', fontsize=22, labelpad=10)
ax_acc.set_ylim(0.6, 1.0)
ax_acc.set_xticks([1, 2, 3, 4, 5])
ax_acc.set_xlim(0.5, 5.5)
ax_acc.set_xticklabels(['90H', '70H', '50/50', '70F', '90F'], fontsize=18)
ax_acc.set_yticks(np.arange(0.6, 1.05, .05))
ax_acc.set_yticklabels(np.arange(0.6, 1.05, .05), fontsize=18)
ax_acc.set_ylabel('Proportion Correct', fontsize=22, labelpad=14)
ax_acc.set_xlabel('Prior Probability Cue', fontsize=22, labelpad=10)
#save figures
if save:
fig_rt.savefig(RTname+'.png', dpi=600)
fig_acc.savefig(ACCname+'.png', dpi=600)
def predict_from_simdfs(data, simdfs, save=True, mname='EvT'):
"""
Arguments:
data (pandas df): pandas dataframe with the empirical data used
to fit the model to generate the simulation parameters
simdfs (pandas df): pandas dataframe with multiple simulated datasets
*plot behavioral data against model predictions from multiple simulations
*If save=True, will save RT and ACC plots to working dir
"""
sns.set_style("white")
if len(data.cue.unique())==3:
x=np.array([1,2,3])
code_type='HNL'
else:
x=np.array([1, 2, 3, 4, 5])
code_type='AllP'
face_acc, house_acc, face_rt, house_rt=parse.get_empirical_means(data=data, code_type=code_type)
sem_list=parse.get_emp_SE(data, code_type)
#init sep figure, axes for RT & ACC data
fig_rt, ax_rt=plt.subplots(1)
sns.despine()
fig_acc, ax_acc=plt.subplots(1)
sns.despine()
fig_acc.subplots_adjust(top=0.9, left=0.15, right=0.88, bottom=0.15)
fig_rt.subplots_adjust(top=0.9, left=0.15, right=0.88, bottom=0.15)
flast_rt=np.zeros([5])
hlast_rt=np.zeros([5])
flast_acc=np.zeros([5])
hlast_acc=np.zeros([5])
for simn, rest in simdfs.groupby('sim_num'):
Ftheo_acc, Htheo_acc = parse.get_theo_acc(simdf=rest, code_type=code_type)
flast_acc, hlast_acc = pred_accPLOT(code_type=code_type, acc_ax=ax_acc, xacc=x, yaccFace=Ftheo_acc, yaccHouse=Htheo_acc, ind=simn, flast_acc=flast_acc, hlast_acc=hlast_acc, mname=mname)
Ftheo_rt, Htheo_rt=parse.get_theo_rt(simdf=rest, code_type=code_type)
flast_rt, hlast_rt = pred_rtPLOT(code_type=code_type, rt_ax=ax_rt, xrt=x, yrtFace=Ftheo_rt, yrtHouse=Htheo_rt, ind=simn, flast_rt=flast_rt, hlast_rt=hlast_rt, mname=mname)
#plot empirical ACC
f_emp_acc=ax_acc.errorbar(x, face_acc, yerr=sem_list[0], elinewidth=2.5, ecolor='k', color='Blue', lw=4.0)
h_emp_acc=ax_acc.errorbar(x, house_acc, yerr=sem_list[1], elinewidth=2.5, ecolor='k', color='Red', lw=4.0)
if mname=='pbm':
ax_acc.legend((ax_acc.lines[-4], ax_acc.lines[-1], ax_acc.lines[0], ax_acc.lines[1]), ('Face Data', 'House Data', 'Face Model', 'House Model'), loc=0, fontsize=18)
sns.despine()
#plot empirical RT
f_emp_rt=ax_rt.errorbar(x, face_rt, yerr=sem_list[2], elinewidth=2.5, ecolor='k', color='Blue', lw=4.0)
h_emp_rt=ax_rt.errorbar(x, house_rt, yerr=sem_list[3], elinewidth=2.5, ecolor='k', color='Red', lw=4.0)
if mname=='pbm':
ax_rt.legend((ax_rt.lines[-4], ax_rt.lines[-1], ax_rt.lines[0], ax_rt.lines[1]), ('Face Data', 'House Data', 'Face Model', 'House Model'), loc=0, fontsize=18)
sns.despine()
flist=[fig_rt, fig_acc]
if save:
fig_rt.savefig(mname+'_rt.png', dpi=400)
fig_acc.savefig(mname+'_acc.png', dpi=400)
