Explore IFRS 9 PIT PD modelling fundamentals, including the three-stage impairment, PD vs TTC, and the role of PIT PD in expected credit loss calculations.

Explore IFRS 9 point-in-time PD modeling for forward-looking credit provisions. Link PD to LGD, EAD, and ECL using macroeconomic data, scenario analysis, Bayesian or Vasicek methods, and SAS or Python.

Discover the end-to-end IFRS 9 PIT PD model development and ECL in SAS, covering data preparation, feature engineering, logistic regression, validation, calibration, and ECL calculation.

Develop a point-in-time ifrs 9 probability-of-default model from a small dataset, using development and validation splits, feature prep, and logistic regression for 12-month default predictions evaluated by auc and calibration.

Load and verify a dataset of 5000 observations for IFRS nine point in time PD modeling by importing CSV, enabling graphics, and inspecting variables with proc contents.

Validate the 5000 loan level PD data with SAS procedures to ensure complete, consistent inputs for IFRS 9 modeling, including PD, staging, and ECL, with forward-looking macro variables.

Review data quality checks confirming 5000 observations across 15 variables, showing a 70/30 split with balanced default rates, enabling robust PD and ECL modeling under IFRS nine.

Develop feature engineering for IFRS nine credit risk models using dev and validation prep datasets, applying score reversal, util log, days past due cap, age squared, and one hot encoding.

Estimate a logistic regression model for default probability using IFRS 9 aligned drivers, including region, employment status, score_rev, dpd_clip, age and age2, with development and validation scoring.

Demonstrates how to evaluate a logistic regression credit risk model with AUC and K-s discrimination metrics, using development and validation samples to ensure IFRS nine readiness.

Assess calibration by comparing predicted PDs with observed default rates under IFRS 9 using decile groups and mean PD and observed default per decile; prepare PD for ECL calculations.

Demonstrate calculating expected credit loss under IFRS 9 by modeling ECL as PD times LGD times EAD in SAS, illustrating zero loss when PD=0 and about 394 at PD=0.98.

Master the end-to-end IFRS 9 probability of default workflow, from data preparation and feature engineering to model validation and regulatory governance.

Prepare robust PD models by gathering loan book and macroeconomic data, cleaning, validating, and selecting variables, then creating default flags and aligning observation and performance windows under IFRS 9.

Explore observation and performance windows in credit risk modeling, learn their sequencing to prevent data leakage, and align with IFRS 9 and Basel guidelines for accurate, compliant PD models.

Understand how observation window and performance window work together in credit risk modeling, designing non-overlapping periods to prevent data leakage and meet IFRS 9 requirements.

Track loan cohorts by origination vintage to monitor credit quality and detect patterns using metrics like default and loss rates, supported by vintage matrices and IFRS nine expected credit loss.

Explore wholesale variables for IFRS 9 PD modeling, integrating borrower characteristics, financial ratios, facility and relationship features, macroeconomic indicators, and qualitative judgments to produce forward-looking credit risk estimates in SAS.

Discover retail credit risk modelling variables that drive PD, LGD, and EAD for mortgages, cards, and personal loans, including demographic, behavioral, credit bureau, macroeconomic, and derived variables.

Balance minority defaults in credit risk models using oversampling, improving recall with methods like random oversampling and SMOTE, while recalibrating probabilities for IFRS 9 and Basel.

Improve IFRS 9 expected credit loss accuracy by enforcing six data quality dimensions: accuracy, completeness, consistency, timeliness, validity, and uniqueness through automated checks, data lineage, continuous monitoring, and governance.

Learn to validate, clean, and prepare IFRS nine data sets in SAS, using inventory checks, deduplication, profiling, missingness audits, and sanity rules for regulator-ready credit risk models.

Impute missing data in IFRS nine models using simple, segment-based, and model-based strategies to preserve relationships, quantify uncertainty, monitor drift with the PSI, cluster variables, and automate governance-ready SAS reports.

Identify why missing data matters in credit risk modeling and apply robust handling, including MCAR, MAR, and not at random, to support accurate PD estimates and IFRS nine compliance.

Master missing data handling in proc logistic with imputation, missing flags, and weight of evidence binning. Follow a best-practice workflow to preserve observations and enable scoring.

Explore missing data treatments for financial modeling under IFRS nine, including categories, flags, hot deck, regression, expert, and cluster-based imputations, emphasizing transparency, documentation, and robust risk predictions.

Demonstrates imputing categorical missing values as explicit levels in sas using coalesce and strip, with freq output and IFRS nine model development context.

Generate binary missing indicators for five fields—age, credit utilization, internal score, employment status, and marital status—preserving the missing signal for imputation, monitoring flags over time in IFRS 9 PD modeling.

Demonstrate mean and median imputation in SAS for age, credit utilization, and internal score, comparing variance, outlier robustness, missing indicators, and IFRS 9 considerations.

Impute categorical missing values using mode in SAS with proc SQL and data step. Demonstrates mode limitations and the use of missing indicators for IFRS nine documentation.

Apply hot deck imputation by matching recipients to donors in strata of employment status, marital status, region, and stage, coalesce missing credit utilization to donor values and validate with means.

Impute missing age values using a regression workflow, splitting data into train and validation sets, fitting a stepwise model with aic, and creating age_imp via coalesce.

Apply k-means clustering to segment borrowers using internal score and credit utilization, impute missing internal scores with cluster means using coalesce, and validate imputation via fast cluster workflow summaries.

Apply SME rules to fill missing categories, creating employment_status_SM and marital_status_SM. Evaluate results with proc freq, assess bias, and document IFRS 9 rationale and stability monitoring.

Enhance IFRS nine modeling by applying a toolkit of missing data treatments, including flags, mean and median imputation, hot deck, regression, clustering, and rule imputations, with an Excel audit pack.

Master data quality and preparation for credit risk analysis, including performance windows, vintage analysis, oversampling techniques, wholesale and retail variables, and imputation in SAS.

Explore survival functions in credit risk modeling and their link to lifetime PD and probability of default. Visualize loan health over time under IFRS 9.

Learn lifetime probability of default modelling for IFRS 9 stage two and stage three ECL, covering cohort survival, transition matrices, and survival analysis with a Cox model, implemented in SAS.

Explore the Cox regression model in SAS for lifetime probability of default, modeling the hazard function with covariates and time varying inputs to derive PD term structures for IFRS 9.

Apply the Cox proportional hazards model to estimate lifetime probability of default, using survival functions and time-varying covariates in SAS, with data prep and model checks.

Apply survival analysis with Cox proportional hazards to model lifetime PD for IFRS 9 ECL, focusing on data engineering, censoring, time-varying covariates, and robust validation.

Define right censoring and establish business rules for events versus non-default exits in credit risk survival analysis, using a SAS pattern to preserve unbiased PD curves.

Explore cohort (vintage) modelling for lifetime probability of default, separating age, calendar, and origination effects; implement data layouts, survival curves, and forward-looking IFRS 9 compliant PDS using macroeconomic scenarios.

Track cohorts that start together to reveal how default risk evolves as loans age, and compare cohorts to separate age, calendar, and cohort effects for lifetime PD modeling.

Explore transition matrices to model how loans move between current, 30 dppd, 60 dppd, and default, and project future credit quality, probability of default, and IFRS 9 lifetime risk.

Leverage SAS proc iml to build lifetime probability of default using transition matrices and repeated matrix multiplication for IFRS 9 compliance, projecting 12-, 24-, and 60-month PD curves.

Calculate the lifetime ECL by breaking the loan into monthly slices and applying PD, LGD, and EAD, discounted to present value under IFRS 9.

Explore how lifetime PD strengthens IFRS 9 ECL and Basel risk frameworks, enabling forward-looking provisions, better pricing, capital planning, early warning, and robust stress testing for deteriorating loans.

Compare point-in-time and lifetime PD concepts, explain IFRS 9 stage transitions from stage one to two and the cliff effect, and show how horizon shifts impact provisions and earnings volatility.

Construct a Cox survival model in SAS to estimate monthly and lifetime probabilities of default from an enriched loan panel, using Kaplan-Meier curves and baseline covariates for IFRS 9 ECL.

Explore cohort analysis for ifrs nine default modeling with SAS, grouping loans by origination month to track defaults and derive Kaplan-Meier survival with 12-month and lifetime PDs for ACL.

Use the transition matrix method to estimate lifetime PD from six states S0–S5 with SAS, enabling forward-looking IFRS 9 PD and ECL calculations.

Master survival analysis for credit risk assessment and forward-looking pd modeling, using data preparation, Kaplan-Meier curves, Cox models, and a transition matrix with macroeconomic covariates.

Examine LGD as the severity of loss at default in the IFRS 9 framework. Identify how collateralization, loan type, seniority, borrower quality, and macroeconomic environment influence LGD and ECL.

Define the LGD dependent variable and construct observation and performance windows while addressing data quality, missing values, outliers, and segmentation for IFRS nine and Basel 3.1 compliant LGD models.

Apply linear regression to model loss given default (LGD), interpreting coefficients for factors like loan to value ratio, collateral coverage, and GDP growth, with SAS or Python implementations and validation.

Explore beta regression and fractional response regression to model LGD between zero and one, using logit links, maximum likelihood, and marginal effects for IFRS nine and ECL.

Inflated beta regression extends beta models by adding zero and one mass, forming a three-part LGD distribution with pi zero, pi one, and a beta middle, for improved risk insight.

Explore how the mixed effects LGD model combines fixed and random effects to capture unobserved factors, sector clustering, and macroeconomic shifts for forward-looking scenario-based LGD forecasts.

Compare four lgd modeling approaches: linear, beta, inflated beta, and mixed effects, and validate them for accuracy, stability, and interpretability under IFRS 9 and Basel III.

Bridge LGD across IFRS 9 and Basel 3.1 by using scenario-weighted, downturn-calibrated LGD to drive ECL provisions and RWA, with governance and validation.

Explore an end-to-end lgd modelling case study turning 3,000 defaulted facilities and macroeconomic data into regulatory-grade lgd models with sas and mixed effects for IFRS 9.

Recaps LGD fundamentals, model evolution from linear to mixed effects, and integration with IFRS 9 and Basel, then transitions to ECL via connected EAD and PD under macroeconomic scenarios.

Apply a logit transformation to LGD to linearize relationships and enable reliable modeling. Demonstrate an OLS model in SAS using LTV, collateral, guarantees, and seniority for IFRS 9 ECL contexts.

Apply beta regression with a logit link in SAS PROC GLIMMIX to estimate LGD, with 70% train split (seed 12345), and score via PLM; evaluate with MAE and RMSE.

Estimate the median lgd using quantile regression to robustly model outliers and asymmetric losses, with a sas workflow that trains on 70/30 data and exports predictions for IFRS nine ecl.

Leverage a nonlinear loss given default model with a decision tree using variance-based splits in SAS, incorporating LTV, collateral type, seniority, region, and ECL for robust risk prediction.

Use simple linear regression in SAS to estimate LGD with LTV, interest rate, days past due, and a guarantee flag. Clip predictions to 0–1 and evaluate with R-squared and RMSE.

Explore a fractional logit LGD model with SAS proc mixed and logit link, using Papke and Wooldridge 1996, a 70/30 split with seed 12345, and evaluation by MAE and RMSE.

Shows beta regression with random regional intercepts using SAS proc glimmix for LGD values between 0 and 1, with LTV, guarantee flag, collateral type, and seniority.

Extend LGD modeling with a two-part inflated beta approach, using a multinomial logistic model for zero and one, plus a beta regression for the mid range to predict LGD.

Explore exposure at default modeling (EAD) as the amount at risk at default. See how drawn balances, expected drawdowns, and CCF drive EAD forecasts under IFRS 9 and Basel.

Explore the credit conversion factor (ccf) as the bridge from undrawn commitments to drawn exposure, and learn how ccf under Basel 3.1 and IFRS 9 drives EAD, provisions, and capital.

Design a regulator-ready data set for EAD modeling, aligning with IFRS 9 and Basel 3.1, by defining observation and default windows, computing CCF and EAD, and ensuring data quality.

Build a transparent baseline for EAD estimation using linear and log linear regression, enabling positive, interpretable predictions and robust validation ahead of Tobit and fractional regression.

Apply Tobit and truncated regression to EAD modeling, respecting natural bounds 0–1, implement in SAS and Python, and interpret results for IFRS nine and Basel 3.1 compliance.

Learn beta and fractional response regression for EAD and CCF, delivering bounded 0–1 predictions, interpretation of mu and phi, and SAS and Python implementations under IFRS 9 and Basel 3.1.

Explore credit conversion factor (CCF) modeling with regression and two-stage logistic approaches, linking utilization, rating, tenor, collateral, and GDP growth to EAD and IFRS 9.

Validate and backtest EAD models across portfolios and cycles, using IFRS 9 and Basel 3.1 lenses to prove accuracy, stability, and governance through data integrity, empirical performance, and ongoing monitoring.

Explore how EAD integrates with IFRS 9 and Basel 3.1, aligning PD, LGD, and CCF to drive provisioning, capital adequacy, and unified risk reporting.

Transform raw exposure data into a fully validated, IFRS 9 compliant end-to-end EAD model that integrates data preparation, estimation, validation, and ECL linkage.

Recap how exposure at default (EAD) drives the ECL framework by linking PD, LGD, and forward-looking EDI under IFRS 9 and Basel 3.1, with governance and scenario analysis.

Prepare IFRS nine PD data by cleaning, transforming, and validating loan and borrower data from internal and external sources, and apply completeness and accuracy quality checks.

Implement continuous model validation and monitoring under IFRS nine to ensure accurate ECL estimates, robust governance, and proactive risk management through calibration, backtesting, and macro scenario updates.

Evaluate credit risk modelling performance through a multifaceted framework using confusion matrix, auc, gini, lift charts, d statistic, and psi to ensure calibration and temporal stability.

Explore the Hosmer-lemeshow test for model calibration in IFRS 9, validating predicted probabilities of default (PD) against observed defaults to support accurate expected credit loss calculations and stage allocation.

Explore the brier score, a measure of probabilistic prediction accuracy used in IFRS nine compliance for credit risk modeling, and learn its calculation, interpretation, calibration, and refinement.

Analyze the k-s statistic to measure model discrimination in credit risk, calculating and interpreting the maximum separation between defaults and non-defaulters for IFRS nine compliance.

Explore the ROC curve, AUC, and C-statistic for IFRS nine credit risk, learn how they measure discrimination between defaulters and non-defaulters, and implement in SAS.

Explore the Gini coefficient, lift curve, and Gaines chart to assess a credit risk model's discrimination, linking Gini to AUC and applying IFRS 9 PD and ECL validation.

Sort accounts by predicted default, divide into ten deciles, and evaluate how defaults concentrate in top deciles to validate model power and guide lending thresholds and IFRS nine validation.

Compare AIC, BIC, and negative two log likelihood to balance fit, complexity in credit risk modeling. Apply the framework to nested models and regulatory needs, balancing prediction and simplicity.

Master the D statistic, or somer's d, as a rank correlation between predicted probabilities and default versus non-default outcomes, and relate it to concordance, discordance, and the AUC.

Discover how the population stability index measures model stability and data drift in credit risk models under IFRS 9. Calculate PSI across bins and interpret thresholds to trigger recalibration.

Apply validation reruns and scoring to logistic regression credit risk models under IFRS 9, ensuring predictive consistency, detecting overfitting, and meeting regulatory scrutiny with stable performance.

Assess IFRS nine model performance using discrimination metrics (AUC, Gini, KS, D), calibration (Hosmer-Lemeshow, Brier), and stability and selection criteria (PSI, AIC/BIC).

Explore cross validation techniques for IFRS 9 credit risk models, including k-fold and time-series methods, to prevent overfitting, ensure out-of-sample stability, and satisfy governance and regulatory backtesting.

Import the IFRS nine dataset from csv, verify 5,000 observations and 16 variables, then stratify a 70/30 train-validation split with seed 20250826 for logistic regression and ROC, KS, Gini evaluation.

Train a logistic regression on the training set to predict default using internal score, age, credit utilization, and DPD, with ROC visualization and internal score as the strongest negative predictor.

Explore model discrimination using ROC and AUC, interpret the Gini coefficient, and evaluate a logistic regression baseline with an AUC around 0.61 and stable performance.

Apply the Kolmogorov-Smirnov statistic to measure model discrimination using decile-based bads and goods in SAS, with training and validation K-S around 0.17 and lift/gains charts to assess ranking.

Evaluate logistic regression calibration with the Hosmer-Lemeshow test and visualize observed versus expected defaults across deciles; report chi-square 10.17 (p=0.2533) and Brier score 0.126.

Explore association statistics in logistic regression, showing concordant and discordant pairs reveal how well the model ranks default risk with validation data using internal score, age, credit utilization, and DPD.

Learn how the population stability index (psi) measures shifts between training and validation distributions to monitor credit risk models, detect drift, and support governance with a unified monitoring framework.

Evaluate credit risk models beyond development, balancing discrimination, calibration plots, and business impact for IFRS 9, using ROC AUC, Gini, and Brier score to inform ECL decisions.

Master the IFRS 9 ECL framework by calculating 12-month and lifetime ECLs from PD, LGD, and EAD, discounting to present value, and producing portfolio- and stage-level regulatory reports with sensitivity analyses.

Explore how IFRS nine ecl ties to prudential expected loss, and master the data lineage, reconciliations, and governance needed for financial statements, pillar three, and regulatory reporting.

Explain the 12-month expected credit loss under IFRS 9, focusing on stage one assets and the PD times LGD times EAD times DF formula, with a practical example.

Explore how IFRS nine impairment uses forward-looking expected credit losses, driven by PD, LGD, and EAD across three stages, with regulatory capital implications.

Master the lifetime expected credit loss (ECL) framework under IFRS 9, applying the three-stage approach and multi-period ECL formula across instruments with forward-looking macro scenarios.

Explore advanced portfolio aggregation for IFRS nine ECL models, weighting PD, LGD, and EAD by exposure to reveal stage-based risk, trends, and concentration insights.

discover the ifrs nine classification pillar, detailing how assets and liabilities are categorized and measured using the business model and spi tests to determine amortized cost, fvoci, or fvtpl.

Explore lifetime expected credit losses under IFRS 9, contrast with 12-month ECL, and apply the multi-period PD, LGD, EAD, and discounting framework.

Explore IFRS 9 regulatory reporting and disclosures for ECL models, covering PD, LGD, EAD, staging, and macroeconomic scenarios to communicate transparent results to regulators, auditors, and investors.

Master the IFRS 9 ECL model by combining PD, LGD, and EAD to compute 12-month and lifetime ECL with discounting, exploring upside and downside scenarios, for regulatory grade reporting.

Explore sensitivity analysis in credit risk models and transparent ECL disclosures under IFRS 7 and IFRS 9, using scenarios, single-factor and weighted approaches for PD, LGD, and EAD.

Demonstrate end-to-end ECL calculation under IFRS 9 by merging PD, LGD, and EAD with discount factors in SAS across stage 1–3 loans, then aggregate by product and apply scenario weighting.

Bridge the structural Merton framework with IFRS 9 pit PD requirements to map market implied PD to 12-month and lifetime PDS, with governance and scenario overlays for listed corporates.

Explain IFRS 9 staging and significant credit risk increase, moving from stage one to stage two, with 12-month vs lifetime ECL, Basel IRB overlays using PD, LGD, and EAD.

Evaluate and validate IFRS nine staging to ensure stage allocations reflect credit risk, with default rate by stage, transition dynamics, lead time to default, sicr triggers, and portfolio share stability.

We walk through a pit pd evaluation workflow in SAS for IFRS 9, covering setup, data import, and initial staging with 12-month proxy bad rates.

Assess timeliness of staging under IFRS 9 by creating a proxy bad flag from defaults or 90 days past due, then summarize by stage to confirm stage 3 defaults.

Leverage the Kolmogorov Smirnov statistic to measure model discrimination in SAS, using deciles, bads and goods, and visualize with lift and gains insights.

Connect the Cox survival outputs to IFRS 9 ecl by deriving marginal and cumulative pd, estimating ead and lgd, and discounting to present value for lifetime and 12-month comparison.

Develop a SAS IFRS 9 staging macro to assign accounts to stage one, two, or three with rule-based labels, then demonstrate PSI drift across cohorts using decile binning.

Explore IFRS 9 forward-looking information in expected credit loss calculations, using macroeconomic scenarios and probability weighting to adjust PD estimates. Learn scenario definitions, weights, and presenting forecast results for executives.

Explore IFRS 9 staging under Basel overlay, detailing stage one, two, and three and their interaction with PD models, 12-month and lifetime ECL, and regulatory governance.

Calibrate point-in-time PD models to reflect current and forecasted economic conditions, using forward-looking adjustments, scenarios, and IFRS 9 requirements to produce realistic expected credit losses.

Weight scenarios by assigning base, upside, and downside probabilities summing to 100% to derive a probability-weighted PD for IFRS 9 compliant expected credit losses.

Implement pit PD calibration and forward-looking adjustments in SAS. Build baseline PD with proc logistic, calibrate with logit intercept shift, model macroeconomic drivers, and automate with SAS macros.

Calibrate point-in-time 12-month PD models in SAS using calibration in the large and logistic scaling to align predicted defaults with observed rates, and export IFRS 9 compliant metrics.

Explore calibration in credit risk models, comparing SIL (calibration in the large) and Platt scaling to adjust raw PD predictions for level and spread, preserving ranking where appropriate.

Master section five calibration and forward looking adjustments through hands-on practice with the provided data set and code scripts, building practical skills for stronger credit risk predictions.

Calibrate the 12-month PD using sill and Platt scaling to adjust logits. Examine data prep, logit calculation, clipping, and evaluation with AIC, likelihood, odds ratios, and the C statistic.

Forecast macro inputs for credit risk using SAS arima models on GDP, unemployment, and rates, then build baseline, upside, and downside scenarios for pd and ecl under IFRS 9.

Overlay macroeconomic drivers into the probability of default using logistic regression to produce forward-looking, IFRS 9 compliant PD paths and scenario-weighted PDs.

Load data step by step in SAS to perform an end-to-end ACL calculation from PD, LGD, EAD to ECL, including CSV import, data cleaning, and validating observation and performance windows.

Clean a dataset with data quality corrections, define the 12-month forward default target, and assess completeness with proc freq and proc means, preparing the data for PD modeling.

Split data chronologically into training, validation, and test sets using time-based cutoffs to prevent look-ahead bias and ensure Basel 3.1 compliant PD modeling, with class balance checked across splits.

This lecture demonstrates feature engineering. Transform predictors into variables using log transforms, outlier cap, seasoning indicators, and a reusable macro across train, validation, and test sets for delinquency prediction.

Build a 12-month probability of default model with logistic regression in proc logistic, identifying utilization score, days past due, and unemployment as key drivers, and assess convergence, fit, and calibration.

Evaluate a 12-month pd model with roc auc and ks on unseen data, using logistic regression; achieves 0.75 auc and ~0.4 ks for regulatory validation.

Perform a diagnostic calibration check for the probability of default (PD) across twelve months, comparing observed defaults to predicted rates in deciles.

Calculate the 12-month expected credit losses using PD, LGD, and EAD, with LGD normalization and data quality checks, yielding a clean portfolio ECL that supports Basel 3.1 governance.