Monte Carlo Results Spreadsheet
The Monte Carlo Results Spreadsheet can be created from the Quick tab of the Monte Carlo Results dialog box. It contains an extensive summary of the outcome of the Monte Carlo experiment. The spreadsheet stores the results with the replications in rows, and the data for each replication in columns. Which data are stored depends on the options selected in the Monte Carlo Analysis dialog box.
As a Monte Carlo experiment is performed, information about the analysis results for each Monte Carlo replication is stored in memory, and is displayed in an overall results spreadsheet at the end of the experiment. Results for each replication are stored in a new row of the spreadsheet. The basic results, stored for all Monte Carlo experiments, include the Monte Carlo seeds used for each replication, the discrepancy function value, the number of iterations required, and a variety of indicators used to determine whether iteration converged satisfactorily.
Following are the codes for the variable names:
- SEED1
- This is the first of the two Monte Carlo seeds.
- SEED2
- This is the second seed, used only in Contaminated Normal distribution generation.
- TERMCODE
- This is the termination code for the analysis. If this is zero, the analysis apparently converged normally. If not, then the following codes apply.
- The relative function change criterion was below the criterion value. This can occur when the function has stabilized, but the gradient and relative cosine criteria do not go to zero, because one of the parameters is on a boundary value.
- The line search algorithm was unable to reduce the discrepancy function along the searched direction.
- The number of iterations reached the maximum permissible value. If necessary, this value may be altered in the Analysis Parameters dialog box.
- Singular covariance matrix was encountered during iteration. On occasion, the parameters will be changed to values that yield a singular estimated covariance matrix. When this happens in maximum likelihood estimation, the discrepancy function cannot be evaluated, so iteration is stopped.
- (this value is currently not in use)
- The iteration was terminated by user request, i.e., the user stopped iteration with the ESC key or the Cancel button.
- DISCREP
- The value of the discrepancy function after iteration.
- RCOS
- The maximum residual cosine criterion.
- GRADIENT
- The maximum absolute value of the gradient elements after iteration.
- NUM_ITER
- The number of iterations required before termination.
- ICSC
- The ICSF invariance criterion.
- ICS
- The ICS invariance criterion.
- RED_PAR
- The number of redundant parameters.
- RED_CON
- The number of redundant constraints.
- BOUNDARY
- The number of active inequality constraints (NAIC), or "boundary cases," after iteration.
- CHI_SQR
- The Chi-square goodness-of-fit statistic.
- DF
- The number of degrees of freedom for the Chi-square statistic.
- PLEVEL
- The probability level for the Chi-square statistic.
- PAR_#
- The parameter values, numbered as they are in the PATH1 analysis syntax. So, for example, PAR_23 is the value for the free parameter numbered 23 in the analysis syntax.
SE_#. The standard errors, numbered in the same way as the parameter numbers.
- RMS_LO
- The lower endpoint of the 90% confidence interval for the Steiger-Lind (1980) RMS index.
- RMS_PT
- The point estimate for the Steiger-Lind (1980) RMS index.
- RMS_HI
- The upper endpoint of the 90% confidence interval for the Steiger-Lind (1980) RMS index.
- NCP_LO
- The lower endpoint of the 90% confidence interval for the population discrepancy function.
- NCP_PT
- The point estimate for the population discrepancy function.
- NCP_HI
- The upper endpoint of the 90% confidence interval for the population discrepancy function.
- AIC
- The rescaled Akaike information criterion.
- BIC
- The Schwarz Bayesian criterion.
- BR_CUD
- The Browne-Cudeck single sample cross-validation index.
- GAMMA_LO
- The upper endpoint of the 90% confidence interval for the population gamma index.
- GAMMA_PT
- The point estimate for the population gamma index.
- GAMMA_HI
- The upper endpoint of the 90% confidence interval for the population gamma index.
- GAMAD_LO
- The upper endpoint of the 90% confidence interval for the adjusted population gamma index.
- GAMAD_PT
- The point estimate for the adjusted population gamma index.
- GAMAD_HI
- The upper endpoint of the 90% confidence interval for the adjusted population gamma index.
- IRGLS
- The IRGLS discrepancy function, if maximum likelihood estimates were obtained.