In TreeAge Pro you can create distributions to be used for Probabilistic Sensitivity Analysis (PSA) which Resample per Expected Value and for Trial Characteristics which Resample per Individual Trial.
To gain confidence with how distributions are sampled, we created the attached model (DistributionBehaviorTest.trex) to demonstrate Resampling per Individual Trial. The model has:
- Three example strategies, each one demonstrating a different way of referencing the distribution.
- One distribution dist_Normal which is re-sampled per Individual Trial.
- The distribution is referenced at multiple places in the model as:
- the value of the state reward for _stage = 1; and
- the value of the transition reward for _stage = 1.
The trial level distributions are only sampled once for each trial – so let’s demonstrate that the value sampled from the distribution for a given trial will not change. The model sets:
- the state reward (Sample_at_A) to the distribution sample for _stage = 1; and
- the transition reward equal to the distribution sample for _stage = 0.
Consider the Monte Carlo Statistics (Microsimulation output) and the All Data report (attached as All Data Report - same samples.pdf). For each trial we can see the cost of each transition reward and state reward are equal to the distribution sample (Dist_1).
Note there is a function called distForce(index), that forces a new sample to be generated every time the distribution is referenced/called. You can substitute it in the attached model in one of the strategies to see how it would change the results.
This demonstration should give you confidence TreeAge Pro is sampling a value from the distribution for each trial in a microsimulation.
To reference the sample from the distribution you can use dist(index) or distributionName or in TreeAge Pro 2017 R2.0 dist("distributionName"). They are all equivalent and give you the same result.