Next: States, Previous: Simulation, Up: Top

Note: The estimation process described below is being implemented in IEST for a single model sub-process. For a multi-process model, repeat the estimation process for each mode sub-process.

Estimation of model coefficients creates a simulation project that holds the model with the estimated parameters as initialization rules.

This process estimates a specified model's parameters from estimates provided from the literature.

- Define States in the model to be estimated.
- Set up Parameters to used in estimation, including estimation coefficients.
- Set up the Studies / Model to be used.
- Set up the Transitions for the studies/models.
- Set up Populations to be used in estimation.
- Double click 'Add New Project' in the main window.
- In the 'Create New Project' window, select 'Estimation', and click OK.
- Now, in the Project Definition form, give a name for the Estimation project (A).
- Select a Study/Model from the table in the bottom left (C) and associated population information from the table in the bottom right (D). These will provide estimation information. Click the up arrow (E). The entry will now appear in the Study/Model table (B). Note that exactly one Model is required and the studies should provide sufficient information to estimate the model coefficient parameters.
- Repeat the previous step for as many studies as needed. To remove a Study/Model, highlight it in the table (B) and click the down arrow (F).
- Set default initial guesses. To set this and other estimation parameters, select the Initial Guess tab (G) and the following view will appear. Then follow the following instructions.
- To add a line to the initial guesses list (H), write the parameter vector in (I), write the values vector (J). Then add the line by pressing button (K). Each line in the initial guesses list (H) should contain a vector of the form [ParameterName1, ParameterName2,...] in the parameter names and a corresponding initial values vector in the parameter values vector that will provide an initial guess for these coefficients. Each line provides a different initial guess that the system will try to use during optimization.
Parameter names can include:
- Coefficient parameter names
- The vector can also start with the reserved word
*AllCoefficients*that includes a value for all the coefficients used by the model to be estimated - System options to guide the optimization process. In most cases, it is recommended not to change these. Also, a user can globally access these parameters through the parameters form. Note that setting a system option parameter in several lines of initial guesses will cause only the last occurrence to be effective for all initial guesses. One assignment to a system option will affect the current initial guess line and all future assignments. See Parameters for a complete list of system option parameters associated with estimation.

- To delete a line from the initial guesses list (H), select it by pressing on it. Then delete the line by pressing button (L).
- Click Save. The form can now be closed. This will trigger validity checking of the data entered and if no error message is displayed, then the data has been saved to memory. Note that the information is not yet saved to a file.

In the main project form titled: Project Definition, select the Estimation Project Tab (M). Then click on 'Optimize Likelihood and Calculate Model Probabilities' (N). The estimation process will start and may take a while to complete. Upon completion, a simulation project will be created and it will contain the estimated model probabilities. The simulation project created will use the same model and create a default population set that requires modification by the user. The estimated coefficients are initialized in stage 0 using the result obtained by the estimation parameter. Note that during estimation you can see the calculation printouts being displayed on the shell window. Also note that the likelihood expressions are dumped as text to the temp directory if ever an analysis is needed.