Результаты (
английский) 3:
[копия]Скопировано!
The test translation. The text 1.
analysis sensitivity and accuracy of the credit portfolio model
1.1 . With the implementation of the calculations in the model portfolio, developed in accordance with this procedure,Is used a number of external input parameters that have an impact on the results of the calculations in the model:
the average through-the-cycle PD portfolio
- Indicators and
The level correlation between companies in the industry (for the segment corporate borrowers, the formulas (8) and (10) of the п.3.4.5) or network segment (for the retail requirements, the formulas of h. 4.5.3 )
The correlation between residue on magazines (for the segment corporate borrowers, p. 3.4.4 )
- correlation macroeconomic parameters
EV in the calculations on the basis simulation requires generating random scenarios,That, in turn, makes a error in the final results. Contribution to the error in the final result fortuitous (it depends on the quality of the input data model),But it can be is minimized by increasing the number of simulation. Thus, an additional parameter, the impact which the model results it is necessary to assess,Is the number of simulations in the calculation method based on Monte Carlo simulation.
To assess the extent of the impact these parameters (number of simulation and the external parameters) to the aggregate the results of the calculations is sensitivity analysis.Sensitivity analysis model is performed by changing a specific parameter values while maintaining the remaining input parameters (if "other things being equal" ) .
The results of a sensitivity analysis and the main conclusions are summarized in the following paragraphs.
1.2 . The number of simulations in the calculation based on simulation Monte Carlo
The number of simulation is specified by the user to initiate the calculation in the model portfolio. When the number of simulations increases the accuracy of estimates, but it increases the time of calculation models.The random errors "tail" estimates derived when симуляциях, is as follows:
1) is carried out recurrence model calculations for a fixed number of simulation;
(2) is calculated from the standard deviation calculation model, obtained in the calculations for a fixed number of simulation;
3) repeating the procedure for a high number of simulation.
This procedure allows you to determine the number of simulations, which is necessary to ensure that accuracy each weekend generating random numbers do not exceed setpoint level.The results repeat calculations for fixed amounts simulation are listed in the following table:
переводится, пожалуйста, подождите..
