PsiConvergence(tol, type, [level], [mean], [stdev], [perc], [val])

PsiConvergence is a property to PsiOutput (only) and is used in partnership with the Psi Statistic, PsiConverged. This property specifies the convergence monitoring information for the output_cell supplied to PsiOutput().

If, after a simulation has been successfully completed, all PsiConverged functions return True, then you might consider lowering the Number of Iterations on the Platform tab of the Analytic Solver Task pane, since there is no need for extra iterations after convergence has been reached. Conversely, if some or all PsiConverged functions are returning False, then you might consider increasing the number of iterations in order to attain convergence for all output functions in the model.

The required argument, tol, indicates the desired positive or negative tolerance amount. This argument must be > 0. The default setting is 0.05.

The required argument, type, specifies the tolerance type entered, either 1 for actual value or 2 for percentage or relative value. The default setting is 2.

Note: The remaining arguments (level, mean, stdev, perc, val)are all optional.

- In Analytic Solver Desktop, convergence can be monitored using any combination of the mean, standard deviation and a specific percentile for the output_cell. When this property is passed using just the two required arguments, tol and type, Analytic Solver performs convergence monitoring using the mean (only) of the output_cell.
- In Analytic Solver Cloud App, RASON Decision Services and Solver SDK, these remaining options exist only for compatibility with @Risk converted models. These options are not utilized in the convergence determination.

The optional argument, level, is ignored in Analytic Solver Desktop and only exists for compatibility with @Risk converted models.

Set the optional argument, mean, to True, in order to use the mean (PsiMean) as a monitoring statistic. If only the first two arguments are passed, tol and type, this option will be automatically set to True.

Set the optional argument, stdev, to True, to use the standard deviation (PsiStdDev) as a monitoring statistic.

Set the optional argument, perc, to True to use a percentile value (specific value passed in the last argument, val) as a monitoring statistic.

If perc is set to True, use the val argument to pass the desired percentile value, from 0.01 to .99.

Example 1:

A1 = PsiNormal(5,6)+PsiOutput("Test", 1, PsiConvergence(0.01, 2))

Where 0.01 is the tol argument and 2 is the type argument, specifying that 0.01 is a percentage.

Example 2:

A1 = PsiNormal(5,6)

A2 = PsiOutput(A1, 1, PsiConvergence(3, 1))

Where 3 is the tol argument and 1 is the type argument, specifying that 3 is a real (actual) value.

Example 3:

A1 = PsiNormal(5,6)

A2 = PsiOutput(A1, 1, PsiConvergence(3, 1, 0, True, True, True, 0.50))

Where 3 is the tol argument and 1 is the type argument, specifying that 3 is a real (actual) value. Three monitoring statistics are turned on, mean (4th argument to PsiConvergence set to True), standard deviation (5th argument to PsiConvergence set to True) and the 50th percentile (6th argument to PsiConvergence set to True and the percentile value passef as 50% in the last argument).

Example 4: Example 1 extended with PsiConverged():

A1=PsiNormal(5,6)+PsiOutput("Test",1,PsiConvergence(0.01,2))

A2 = PsiConverged(A1,2)

Output for A2 will be either True, indicating that the mean (PsiMean) of the output function in A1 has converged to within 1% (.01) of its true value in the 2nd simulation. If not, this function will return False.

See the PsiOutput and PsiConverged definitions for information on their respective arguments.