******** ML Estimation programs 
*** just copy into execution window and press enter ***********

*** total model.

program define ML_choice_time_conf
version 1.0
args lnf p mu0T sigmaT distT  mu0C sigmaC distC
quietly replace `lnf'=ln(binomial(total_pred, $ML_y1,`p')-binomial(total_pred, $ML_y1-1,`p')) if index==1 & miss == 0
quietly replace `lnf'=ln(binomial(total_pred, $ML_y1, 0.5)-binomial(total_pred, $ML_y1-1, 0.5)) if index==1 & miss == 1

quietly replace `lnf'=ln(normalden( $ML_y1, `mu0T'+abs(`distT')* total_pred, `sigmaT' )) if index==2 

quietly replace `lnf'=ln(normalden( $ML_y1, `mu0C'+abs(`distC')* total_pred, `sigmaC' )) if index==3 
end

** Flat time prediction

program define ML_choice_time_conf_flat_time
version 1.0
args lnf p mu0T sigmaT mu0C sigmaC distC
quietly replace `lnf'=ln(binomial(total_pred, $ML_y1,`p')-binomial(total_pred, $ML_y1-1,`p')) if index==1 & miss == 0
quietly replace `lnf'=ln(binomial(total_pred, $ML_y1, 0.5)-binomial(total_pred, $ML_y1-1, 0.5)) if index==1 & miss == 1

quietly replace `lnf'=ln(normalden( $ML_y1, `mu0T', `sigmaT' )) if index==2 

quietly replace `lnf'=ln(normalden( $ML_y1, `mu0C'+abs(`distC')* total_pred, `sigmaC' )) if index==3 
end

** Flat Confidence Prediction

program define ML_choice_time_conf_flat_conf
version 1.0
args lnf p mu0T sigmaT distT  mu0C sigmaC 
quietly replace `lnf'=ln(binomial(total_pred, $ML_y1,`p')-binomial(total_pred, $ML_y1-1,`p')) if index==1 & miss == 0
quietly replace `lnf'=ln(binomial(total_pred, $ML_y1, 0.5)-binomial(total_pred, $ML_y1-1, 0.5)) if index==1 & miss == 1

quietly replace `lnf'=ln(normalden( $ML_y1, `mu0T'+abs(`distT')* total_pred, `sigmaT' )) if index==2 

quietly replace `lnf'=ln(normalden( $ML_y1, `mu0C', `sigmaC' )) if index==3 
end


** Flat time and confidence prediction

program define ML_choice_time_conf_flat_both
version 1.0
args lnf p mu0T sigmaT mu0C sigmaC
quietly replace `lnf'=ln(binomial(total_pred, $ML_y1,`p')-binomial(total_pred, $ML_y1-1,`p')) if index==1 & miss == 0
quietly replace `lnf'=ln(binomial(total_pred, $ML_y1, 0.5)-binomial(total_pred, $ML_y1-1, 0.5)) if index==1 & miss == 1

quietly replace `lnf'=ln(normalden( $ML_y1, `mu0T', `sigmaT' )) if index==2 

quietly replace `lnf'=ln(normalden( $ML_y1, `mu0C', `sigmaC' )) if index==3 
end


** Random Choice Model

program define ML_choice_time_conf_random
version 1.0
args lnf mu0T sigmaT mu0C sigmaC 

quietly replace `lnf'=ln(binomial(total_pred, $ML_y1, 0.5)-binomial(total_pred, $ML_y1-1, 0.5)) if index==1

quietly replace `lnf'=ln(normalden( $ML_y1, `mu0T', `sigmaT' )) if index==2 

quietly replace `lnf'=ln(normalden( $ML_y1, `mu0C', `sigmaC' )) if index==3 
end



****** single estimations

**choices only

program define ML_choice
version 1.0
args lnf p 
quietly replace `lnf'=ln(binomial(total_pred, $ML_y1,`p')-binomial(total_pred, $ML_y1-1,`p')) if index==1 & miss == 0
quietly replace `lnf'=ln(binomial(total_pred, $ML_y1, 0.5)-binomial(total_pred, $ML_y1-1, 0.5)) if index==1 & miss == 1
end


**time only

program define ML_time
version 1.0
args lnf mu0T sigmaT distT

quietly replace `lnf'=ln(normalden( $ML_y1, `mu0T'+`distT'* total_pred, `sigmaT' )) if index==2 

end


**confidence only

program define ML_conf
version 1.0
args lnf mu0C sigmaC distC
quietly replace `lnf'=ln(normalden( $ML_y1, `mu0C'+`distC'* total_pred, `sigmaC' )) if index==3 
end

*********End Definition



Stop

*** Total estimation program for all Participants and all strategies producing a participant x strategy matrix with BIC scores


matrix strat_vec = [0,1,1,0,4]
*** determines which model is used per strategy: 0-full model, 1-flat decision time prediction, 2-flat confidence prediction, 3-flat both, 4-random choice model 

matrix N_obs_cat = [18,18,18,18,18] 
*** replace by number or observations


matrix result = J(10,5,0)


***** replace after the two "/" by total number of participants  and strategies

forvalues u= 1/10 {

forvalues v = 1/5 {

disp "*******************************************"

disp "**** Subject: "`u' "   **** Strategy: "`v'

if strat_vec[1,`v'] ==0 { 

ml model lf ML_choice_time_conf (epsilon: dv=) (mu_Time:) (sigma_Time:) (R_Time:) (mu_Conf:) (sigma_Conf:) (R_Conf:) if Participant ==`u'& strat == `v'
ml init 0.5 0  2500 1000 37  45 30, copy
ml maximize
matrix coef = e(b)
if coef[1,1] < .50 {
 scalar BIC = -2* e(ll)+ln(N_obs_cat[1, `v'])*7
matrix result[`u', `v'] = BIC

disp "Subj: " `u'  "  Strat: " `v' "   - BIC:" BIC 
}
if coef[1,1] >= .50 {
disp "BIC: not calculated because error bigger than .50" 
disp "Subj:  " `u'  "   Strat: " `v' "   - BIC: not calculated because epsilon bigger or equal than .50" 
}
}

if strat_vec[1,`v'] ==1 { 

ml model lf ML_choice_time_conf_flat_time (epsilon: dv=) (mu_Time:) (sigma_Time:) (mu_Conf:) (sigma_Conf:) (R_Conf:) if Participant ==`u'& strat == `v'
ml init 0.5 0  2500  37  45 30, copy
ml maximize
matrix coef = e(b)
if coef[1,1] < .50 {
 scalar BIC = -2* e(ll)+ln(N_obs_cat[1, `v'])*6
matrix result[`u', `v'] = BIC

disp "Subj: " `u'  "  Strat: " `v' "   - BIC:" BIC 
}
if coef[1,1] >= .50 {
disp "BIC: not calculated because error bigger than .50" 
disp "Subj:  " `u'  "   Strat: " `v' "   - BIC: not calculated because epsilon bigger or equal than .50" 
}
}

if strat_vec[1,`v'] ==2 { 

ml model lf ML_choice_time_conf_flat_conf (epsilon: dv=) (mu_Time:) (sigma_Time:) (R_Time:) (mu_Conf:) (sigma_Conf:) if Participant ==`u'& strat == `v'
ml init 0.5 0  2500 1000 37  45, copy
ml maximize
matrix coef = e(b)
if coef[1,1] < .50 {
 scalar BIC = -2* e(ll)+ln(N_obs_cat[1, `v'])*6
matrix result[`u', `v'] = BIC

disp "Subj: " `u'  "  Strat: " `v' "   - BIC:" BIC 
}
if coef[1,1] >= .50 {
disp "BIC: not calculated because error bigger than .50" 
disp "Subj:  " `u'  "   Strat: " `v' "   - BIC: not calculated because epsilon bigger or equal than .50" 
}
}
if strat_vec[1,`v'] ==3 { 

ml model lf ML_choice_time_conf_flat_both (epsilon: dv=) (mu_Time:) (sigma_Time:) (mu_Conf:) (sigma_Conf:) if Participant ==`u'& strat == `v'
ml init 0.5 0  2500 37  45 , copy
ml maximize
matrix coef = e(b)
if coef[1,1] < .50 {
 scalar BIC = -2* e(ll)+ln(N_obs_cat[1, `v'])*5
matrix result[`u', `v'] = BIC

disp "Subj: " `u'  "  Strat: " `v' "   - BIC:" BIC 
}
if coef[1,1] >= .50 {
disp "BIC: not calculated because error bigger than .50" 
disp "Subj:  " `u'  "   Strat: " `v' "   - BIC: not calculated because epsilon bigger or equal than .50" 
}
}
if strat_vec[1,`v'] ==4 { 

ml model lf ML_choice_time_conf_random (mu_Time: dv=) (sigma_Time:) (mu_Conf:) (sigma_Conf:) if Participant ==`u'& strat == 4
ml init 0  2500 37  45, copy
ml maximize
matrix coef = e(b)
scalar BIC = -2* e(ll)+ln(N_obs_cat[1, `v'])*4
matrix result[`u', `v'] = BIC

disp "Subj: " `u'  "  Strat: " `v' "   - BIC:" BIC 
}
}
}

matrix list result