/* Our general hypothesis was that SRAD is characterized by more impulsive decision-making, i.e. steeper delay discounting, increased risk-taking for probabilistic gains, decreased risk-taking for probabilistic losses, and decreased loss aversion. We hypothesized that these different facets of impulsive decision-making (1) differ between individuals with SUD and ND compared to healthy controls. To answer the first research question (group differences), we used Bayesian linear regression with the predictor SRAD (two dummy-coded variables for SUD and ND and the control group as reference) and the logarithmic k-values or λ (for mixed gambles) as outcomes. This was done to achieve more stable estimates with normal distributions for both differences from controls (regression coefficient) and parameters (expectation/ mean and variance) taken from a previous study of our lab [Bernhardt et al. 2017]. In line with common models on SRAD, we assumed no differences between specific forms of SUD or ND. Therefore, we used the group differences from the mentioned previous study that applied the same task battery to compare a sample of individuals with alcohol use disorder with healthy controls as priors. This enabled us to calculate how likely group differences greater (for delay discounting) resp. lower (for the other tasks) than zero (alternative hypothesis) were - given the priors. */ use "data_set.dta" //variable explanation: ddp_log_k_z= z-standardized values of ddp_log_k, group: 1=ND group, 2=SUD group, 3=control group //delay discounting paradigm (ddp) //informativem prior [Bernhardt et al. 2017]: ddp MW 0.37 SE 0.14 Var 0.02 bayes , rseed(64674376) prior({ ddp_log_k_z : 1.group }, normal(0.37,0.02)) mcmcsize(100000) burnin(5000) thinning(2) saving("C:\example_directory\posterior_ddp", replace): regress ddp_log_k_z ib3.group // how likely are group differences greater than zero (alternative hypothesis) bayestest interval {ddp_log_k_z: 1.group}, lower(0) upper(10) bayestest interval {ddp_log_k_z: 2.group}, lower(0) upper(10) //probability discounting for losses (pdl) //informativem prior [Bernhardt et al. 2017]: pdl MW -0.16 SE 0.15 Var 0.02 bayes , rseed(64674376) prior({pdl_kN_z : 1.group }, normal(-0.16,0.02)) mcmcsize(100000) burnin(5000) thinning(2) saving("C:\example_directory\posterior_pdl", replace): regress pdl_kN_z ib3.group // how likely are group differences lower than zero (alternative hypothesis) bayestest interval {pdl_kN_z: 1.group}, lower(-10) upper(0) bayestest interval {pdl_kN_z: 2.group}, lower(-10) upper(0) //probability discounting for wins/ gains (pdw) //informativem prior [Bernhardt et al. 2017]: pdl MW -0.16 SE 0.15 Var 0.02 bayes , rseed(64674376) prior({pdw_kN_z : 1.group }, normal(-0.16,0.02)) mcmcsize(100000) burnin(5000) thinning(2) saving("C:\example_directory\posterior_pdw", replace): regress pdw_kN_z ib3.group // how likely are group differences lower than zero (alternative hypothesis) bayestest interval {pdw_kN_z: 1.group}, lower(-10) upper(0) bayestest interval {pdw_kN_z: 2.group}, lower(-10) upper(0) //mixed gamble (mg) //informativem prior [Bernhardt et al. 2017]: mg MW -0.44 SE 0.14 Var 0.02 bayes , rseed(64674376) prior({mg_lN_z : 1.group }, normal(-0.44,0.02)) mcmcsize(100000) burnin(5000) thinning(2) saving("C:\example_directory\posterior_mg", replace): regress mg_lN_z ib3.group // how likely are group differences lower than zero (alternative hypothesis) bayestest interval {mg_lN_z: 1.group}, lower(-10) upper(0) bayestest interval {mg_lN_z: 2.group}, lower(-10) upper(0) //Finally, the diagnostic groups were compared with the difference in simulated posterior distributions between the SUD-controls and the ND-controls coefficients. //ddp //0 = k-value group SUD > ND, 1 = k-value group SUD < ND use C:\example_directory\posterior_ddp gen diff_ddp=eq1_p1-eq1_p2 gen diff01_ddp=diff_ddp>0 tab diff01_ddp //pdw //0 = k-value group SUD < ND, 1 = k-value group SUD > ND use C:\example_directory\posterior_pdw gen diff_pdw=eq1_p1-eq1_p2 gen diff01_pdw=diff_pdw>0 tab diff01_pdw //pdl //0 = k-value group SUD < ND, 1 = k-value group SUD > ND use C:\example_directory\posterior_pdl gen diff_pdl=eq1_p1-eq1_p2 gen diff01_pdl=diff_pdl>0 tab diff01_pdl //mg //0 = lambda group SUD < ND, 1 = lambda group SUD > ND use C:\example_directory\posterior_mg gen diff_mg=eq1_p1-eq1_p2 gen diff01_mg=diff_mg>0 tab diff01_mg