linushof / sampling-strategies Goto Github PK

• remove distinction between safe and risky
• remove distinction between ev and mean (only mean)
• only one threshold type (put 2nd in appendix)

• use lower choice difficulty (higher EV differences) to reduce sample size and prevent ceiling effects
• problems without rare event produce random more random behavior - consider using (60/40) rather than (50/50) choice problems

• Large chain and iteration number
• alpha in [0,2]
• First weighting for both comparison rules, then value
• remove double facet coding for switching probability

• interaction of sampling strategies with choice problems (impact of rare event)

Description of choice problems are not fully visible

• Plot how sampling strategies influence the proportion of choosing the prospect that offers a higher return most of the time.

• parameter recovery study assuming low values for choice consistency parameter
• posterior predictive check
• doublecheck manuscript

Use separate data sets for relative and absolute thresholds

• remove choice problem features from simulation data set (only use id)
• choice als logical (not character)