Model Properties - zward/Amua GitHub Wiki
Model properties can be accessed from the main menu under Model -> Properties. The following properties tabs are available:
Analysis Properties
Model Dimensions
By default, models in Amua are created with Costs as the outcome of interest, but other dimensions can be easily added. For example, Costs, QALYs, and Deaths could all be used simultaneously in the model. Each dimension should be given a symbol for the model display, and dimensions are separated by a semi-colon ; in the data entry fields. The number of decimal places reported by the model can also be specified here.
Analysis Type
| Analysis Type | Description |
|---|---|
| Expected Value | Calculates the expected value of each model dimension |
| Cost-Effectiveness Analysis (CEA) | Calculates the incremental cost-effectiveness ratio (ICER) of all strategies compared to the specified baseline strategy. Model dimensions must be selected as 'Cost' and 'Effect' |
| Benefit-Cost Analysis (BCA) | Calculates the net monetary benefit (NMB) of all strategies based on the specified willingness-to-pay (WTP) threshold. Model dimensions must be selected as 'Cost' and 'Benefit' |
Simulation Properties
| Property | Description |
|---|---|
| Simulation type | Models can be run as a cohort simulation or as an individual-level Monte Carlo simulation. Selecting Monte Carlo simulation enables other simulation options and the allows for the definition of subgroups (see below) |
| Cohort size/# simulations | Specifies the size of the cohort if cohort simulation is selected, or the number of individuals to simulate if Monte Carlo simulation is selected |
| Seed RNG and Seed | Allows the random number generator to be seeded for Monte Carlo simulation so that the same results are obtained each time the model is run |
| Display individual-level results | Displays a summary of individual-level results from a Monte Carlo simulation.Note: Selecting this option uses more memory and may affect model performance. You may need to increase the Java heap size (see FAQ) if simulating many individuals. |
| Multi-thread simulation | Allows the model to be run on multiple cores.Note: This option will typically always be faster than serial execution for Monte Carlo Decision Trees, but due to extra overhead it may not always result in faster Markov Monte Carlo simulations. Benchmarking the simulation time with different numbers of threads is recommended to ensure the best performance |
Markov Properties
Additional properties can be set for Markov Models .
| Property | Description |
|---|---|
| Max cycles | A backstop used to stop the simulation in case the termination condition is never met |
| Show trace | Option to display the Markov trace or not after simulation |
| Compile traces | Option to compile Markov traces from each chain into one window |
| State Prevalence Decimals | Sets the decimal places for rounding state prevalence in the Markov Trace |
| Half-cycle correction | If selected this will apply a half-cycle correction to the rewards in the first and last cycle of the simulation |
| Discounting | A discount rate (r) can be specified for each model dimension and the first cycle in which discounting should occur can also be specified, as well as the number of cycles per year (n). Discount factor = 1/(1+r)^(t/n) |
Subgroup Properties
Subgroup results can be calculated for Monte Carlo simulations. This is especially useful for Extended Cost-Effectiveness Analysis.
Subgroups are defined based on individual-level variables. Each subgroup must have a unique name, but they do not have to be mutually exclusive. Inclusion in subgroups is determined by boolean expressions and can be based on discrete variables (e.g. urban/rural) or continuous variables (e.g. quantile of SES), or some combination (e.g. urban poor).