Because the gamma distribution created some extremely high numbers we end up redistributin anyone who scored higher than 80 Standardard drinks. These two plots represent the number of mean times this occured at a given level of a simulation and the max times it occurred. The total number of cells in the data set was 80640 therefore even though this occurred it was only for a small number of the total data set.
Simulation Plots
I simulated 6 different levels of mode (range = 1.0 to 1.5) and 10 different levels of SD (range = 0.0 to 0.14)both have an impact on the mean of the drinking bias (X axis). There are 1000 simulations at each combination of levels of mode and SD (each point on the graph) for a total of 90000 simulations.
The black horizontal line on each point represents the value of the original COMBINE data set (N = 896; after removing 487 subjects due to missigness)
Drink Bias Plots
These are profiles of the drink bias distributions we used for each level of the simulation. These are based on a gamma distribution which is parameterized by two different parameters (shape and rate). However, one can convert a mode and SD of a distribution into a shape and rate parameters.
