R: Probability of Default Calibration using Quasi Moment...
QMMRecalibrate
R Documentation
Probability of Default Calibration using Quasi Moment Matching Algorithm
Description
Calibrates conditional probabilities of default (PD) according to Quasi Moment Matching (QMM) algorithm.
Calibration is based on target accuracy ratio (AR) and mean portfolio PD (Central Tendency). For the information purposes, also AR standard deviation is estimated using bootstrap approach.
Target Mean PD (Central Tendency) for the portfolio.
pd.cond.old
Conditional PD distribution.
portf.uncond
Unconditional portfolio distribution.
portf.condND
Conditional on non-default portfolio distribution.
If portf.condND is NULL, portf.uncond will be used as a proxy.
AR.target
Target accuracy ratio(AR), in case is NULL - implied by pd.cond.old AR is used (ARestimate is called for AR estimation purposes).
rating.type
In case 'RATING', each item in the portf.uncond contains number of counterparts in a given rating class.
In case 'SCORE', each item in the portf.uncond is treated as an exact score of counterparty.
calib.curve
In case 'logit', simple logit calibration curve is used (is applicable only for rating.type = 'SCORE').
In case 'robust.logit', robust logit function is used (see Tasche D. (2013) for details).
Details
PD curve is fitted using robust logit function proposed by D. Tasche.
For the information purposes output of the function also contains PD fitted using target CT and AR plus/minus one standard deviation.
Value
alpha
Itercept parameter of the calibration curve.
beta
Slope parameter of the calibration curve.
CT.ac
Mean PD after calibration, e.g. target CT.
AR.ac
AR after calibration, e.g. target AR.
CT.bc
Mean PD before calibration, as implied by conditional PDs and portfolio unconditional distribution.
AR.bc
AR before calibration estimated from conditional PDs.
AR.sdev
AR standard deviation (based on sample data).
condPD.ac
Conditional PDs after QMM calibration.
condPD.bc
Conditional PDs before calibration.
condPD.ac.upper
Conditional PDs given AR as initial AR plus one standard deviation and target CT.
condPD.ac.lower
Conditional PDs given AR as initial AR minus one standard deviation and target CT.
portf.cumdist
Cumulative portfolio distribution needed to estimate logit PDs (conditional on non-default portfolio distribution if such data is given).
portf.uncond
Unconditional portfolio distribution from the worst to the best credit quality.
rating.type
In case 'RATING', each item in the portf.uncond contains number of counterparts in a given rating class.
In case 'SCORE', each item in the portf.uncond is treated as an exact score of counterparty.
Note
Portfolio and default data should be sorted by rating classes from lowest credit quality to higher one.
Author(s)
Denis Surzhko <densur@gmail.com>
References
Tasche, D. (2009) Estimating discriminatory power and PD curves when the number of defaults is small. Working paper, Lloyds Banking Group.
Tasche, D. (2013) The art of probability-of-default curve calibration. Journal of Credit Risk, 9:63-103.
R version 3.3.1 (2016-06-21) -- "Bug in Your Hair"
Copyright (C) 2016 The R Foundation for Statistical Computing
Platform: x86_64-pc-linux-gnu (64-bit)
R is free software and comes with ABSOLUTELY NO WARRANTY.
You are welcome to redistribute it under certain conditions.
Type 'license()' or 'licence()' for distribution details.
R is a collaborative project with many contributors.
Type 'contributors()' for more information and
'citation()' on how to cite R or R packages in publications.
Type 'demo()' for some demos, 'help()' for on-line help, or
'help.start()' for an HTML browser interface to help.
Type 'q()' to quit R.
> library(LDPD)
> png(filename="/home/ddbj/snapshot/RGM3/R_CC/result/LDPD/QMMRecalibrate.Rd_%03d_medium.png", width=480, height=480)
> ### Name: QMMRecalibrate
> ### Title: Probability of Default Calibration using Quasi Moment Matching
> ### Algorithm
> ### Aliases: QMMRecalibrate
> ### Keywords: credit risk probability of default PD calibration
>
> ### ** Examples
>
> pd <- c(0.2, 0.1, 0.005, 0.001, 0.001)
> porfolio <- c(100, 200, 200, 200, 100)
> qmm <- QMMRecalibrate(0.05, pd, porfolio, rating.type = 'RATING')
> QMMPlot(qmm)
>
>
>
>
>
> dev.off()
null device
1
>