• | Maximum number of iteration for the Riccati solver: Control for selecting the maximum number of iterations for the Riccati solver. Integer values from 1,000 up to 25,000 are supported |
• | Tolerance value for the Riccati solver: Control for selecting the tolerance value for the Riccati solver. Values between 0.01 and 0.00000000001 are supported. |
• | Length of impulse response function horizon: Select how many periods to use for impulse responses. Integer values between 1 and 50 years are supported. |
• | Step length for finite difference approximation of inverse Hessian: Lets you select the step length for the finite difference approximation of the inverse Hessian. Values between 0.0001 and 1 are supported. |
• | Confidence region estimation method: Select which confidence region (credible region) estimator YADA should use. The equal tails and highest probability density (HPD) for a unimodal distribution methods are supported. |
• | Check the optimum: Decide if YADA should run the check optimum routines when the posterior mode estimation has finished. The check optimum function can also be run separately from the View Menu through the Check Posterior Mode menu function.. |
• | Check only for transformed parameters: Option to check only for the transformed parameters. If it is unchecked or if the optimization routine targets the original parameters, then YADA will also run the check optimum routine for the original parameters. |
• | Transform conditional standard deviations for modified Hessian to marginal using correlations from Hessian: If this option is check marked and you have opted to for the quadratic approximation to the log posterior for your inverse Hessian estimator, then the conditional standard deviations that are calculated when the check optimum routine is run are transformed to marginal standard deviations by using the correlations from the estimate of the inverse Hessian that the optimization routine provides. |
• | Use fixed state to initialize random number generators: If you check mark this option then YADA will always set the state for the random number generators to 0 before sampling. This guarantees that you can replicate random draws. If unchecked, then YADA sets the state for the random number generators to the sum of 100 times the value obtained from the clock function in Matlab. The clock function provides the vector output: [year month day hour minute seconds]. |
• | Randomize draws from posterior distribution: Whe using a subset of posterior draws, YADA can either choose a common equal distance interval for the draws or select the draws randomly via a uniform distribution. If this option is check marked, then YADA will use the latter selection scheme. See the "Maximum number of posterior draws to use for prediction" and "Percentage use of posterior draws for impulse responses, etc" options on the Posterior Sampling frame of the Options tab. |
• | Compute posterior standard deviations from inverse Hessian using finite differences: If check marked, then YADA will compute an inverse Hessian using finite differences when running the check optimum routine. |

Additional Information
• | A more detailed description about the Riccati solver for forecast error variance decompositions is found in Sections 11.4 and 11.5 of the YADA Manual. |
• | The maximum probability density (MPD) estimator of a confidence/credible region for a unimodal distribution is presented in Section 8.4 of the YADA Manual. |
Page url:
http://www.texlips.net/yada/help/index.html?tools.htm