On Predictive Modeling for Claim Severity Paper in Spring 2005 CAS Forum Glenn Meyers ISO Innovative Analytics Predictive Modeling Seminar September 19,

Slides:



Advertisements
Similar presentations
Tests of Hypotheses Based on a Single Sample
Advertisements

Chapter 9 Introduction to the t-statistic
Copyright (c) 2004 Brooks/Cole, a division of Thomson Learning, Inc. Chapter 9 Inferences Based on Two Samples.
Lecture (11,12) Parameter Estimation of PDF and Fitting a Distribution Function.
Uncertainty and confidence intervals Statistical estimation methods, Finse Friday , 12.45–14.05 Andreas Lindén.
Chap 9: Testing Hypotheses & Assessing Goodness of Fit Section 9.1: INTRODUCTION In section 8.2, we fitted a Poisson dist’n to counts. This chapter will.
What role should probabilistic sensitivity analysis play in SMC decision making? Andrew Briggs, DPhil University of Oxford.
Analysis and Interpretation Inferential Statistics ANOVA
1. Estimation ESTIMATION.
EPIDEMIOLOGY AND BIOSTATISTICS DEPT Esimating Population Value with Hypothesis Testing.
Chapter Seventeen HYPOTHESIS TESTING
Fundamentals of Hypothesis Testing. Identify the Population Assume the population mean TV sets is 3. (Null Hypothesis) REJECT Compute the Sample Mean.
SIMPLE LINEAR REGRESSION
Inference about a Mean Part II
Using ranking and DCE data to value health states on the QALY scale using conventional and Bayesian methods Theresa Cain.
Inferences About Process Quality
SIMPLE LINEAR REGRESSION
Chapter 9 Hypothesis Testing II. Chapter Outline  Introduction  Hypothesis Testing with Sample Means (Large Samples)  Hypothesis Testing with Sample.
Lecture II-2: Probability Review
Commercial Property Size of Loss Distributions Glenn Meyers Insurance Services Office, Inc. Casualty Actuaries in Reinsurance June 15, 2000 Boston, Massachusetts.
Review of normal distribution. Exercise Solution.
SIMPLE LINEAR REGRESSION
Review of Statistical Inference Prepared by Vera Tabakova, East Carolina University ECON 4550 Econometrics Memorial University of Newfoundland.
1 Bayesian methods for parameter estimation and data assimilation with crop models Part 2: Likelihood function and prior distribution David Makowski and.
Statistical Decision Theory
Prof. Dr. S. K. Bhattacharjee Department of Statistics University of Rajshahi.
Random Sampling, Point Estimation and Maximum Likelihood.
Statistics for the Behavioral Sciences Second Edition Chapter 11: The Independent-Samples t Test iClicker Questions Copyright © 2012 by Worth Publishers.
Practical GLM Modeling of Deductibles
The Common Shock Model for Correlations Between Lines of Insurance
 Copyright 2006 National Council on Compensation Insurance, Inc. All Rights Reserved. BAYESIAN ESTIMATION OF STATE SPACE RESERVING MODELS Casualty Loss.
Testing Models on Simulated Data Presented at the Casualty Loss Reserve Seminar September 19, 2008 Glenn Meyers, FCAS, PhD ISO Innovative Analytics.
MEGN 537 – Probabilistic Biomechanics Ch.5 – Determining Distributions and Parameters from Observed Data Anthony J Petrella, PhD.
Estimating the Predictive Distribution for Loss Reserve Models Glenn Meyers Casualty Loss Reserve Seminar September 12, 2006.
Toward a unified approach to fitting loss models Jacques Rioux and Stuart Klugman, for presentation at the IAC, Feb. 9, 2004.
Introduction to Inferential Statistics Statistical analyses are initially divided into: Descriptive Statistics or Inferential Statistics. Descriptive Statistics.
Sampling Error SAMPLING ERROR-SINGLE MEAN The difference between a value (a statistic) computed from a sample and the corresponding value (a parameter)
Goodness-of-Fit Chi-Square Test: 1- Select intervals, k=number of intervals 2- Count number of observations in each interval O i 3- Guess the fitted distribution.
Interval Estimation and Hypothesis Testing Prepared by Vera Tabakova, East Carolina University.
LECTURE 25 THURSDAY, 19 NOVEMBER STA291 Fall
CHEMISTRY ANALYTICAL CHEMISTRY Fall Lecture 6.
Statistical Decision Theory Bayes’ theorem: For discrete events For probability density functions.
Sampling distributions rule of thumb…. Some important points about sample distributions… If we obtain a sample that meets the rules of thumb, then…
Review of Statistical Inference Prepared by Vera Tabakova, East Carolina University ECON 4550 Econometrics Memorial University of Newfoundland.
Inferences Concerning Variances
Stochastic Loss Reserving with the Collective Risk Model Glenn Meyers ISO Innovative Analytics Casualty Loss Reserving Seminar September 18, 2008.
G. Cowan Computing and Statistical Data Analysis / Stat 9 1 Computing and Statistical Data Analysis Stat 9: Parameter Estimation, Limits London Postgraduate.
Statistics Sampling Distributions and Point Estimation of Parameters Contents, figures, and exercises come from the textbook: Applied Statistics and Probability.
Sampling Distribution (a.k.a. “Distribution of Sample Outcomes”) – Based on the laws of probability – “OUTCOMES” = proportions, means, test statistics.
CARe Seminar ILF estimation Oliver Bettis 15 th September 2009.
Estimating the Predictive Distribution for Loss Reserve Models Glenn Meyers ISO Innovative Analytics CAS Annual Meeting November 14, 2007.
Review of Statistical Inference Prepared by Vera Tabakova, East Carolina University.
Parameter Estimation. Statistics Probability specified inferred Steam engine pump “prediction” “estimation”
C HAPTER 2  Hypothesis Testing -Test for one means - Test for two means -Test for one and two proportions.
Statistical Inference for the Mean Objectives: (Chapter 8&9, DeCoursey) -To understand the terms variance and standard error of a sample mean, Null Hypothesis,
MEGN 537 – Probabilistic Biomechanics Ch.5 – Determining Distributions and Parameters from Observed Data Anthony J Petrella, PhD.
 List the characteristics of the F distribution.  Conduct a test of hypothesis to determine whether the variances of two populations are equal.  Discuss.
McGraw-Hill/Irwin © 2003 The McGraw-Hill Companies, Inc.,All Rights Reserved. Part Four ANALYSIS AND PRESENTATION OF DATA.
Can't Type? press F11 Can’t Hear? Check: Speakers, Volume or Re-Enter Seminar Put ? in front of Questions so it is easier to see them. 1 Welcome to Unit.
STA 291 Spring 2010 Lecture 19 Dustin Lueker.
9.3 Hypothesis Tests for Population Proportions
Part Four ANALYSIS AND PRESENTATION OF DATA
Math 4030 – 10b Inferences Concerning Variances: Hypothesis Testing
Inference Concerning a Proportion
More about Posterior Distributions
Interval Estimation and Hypothesis Testing
SIMPLE LINEAR REGRESSION
SIMPLE LINEAR REGRESSION
Chapter 6 Confidence Intervals.
How Confident Are You?.
Presentation transcript:

On Predictive Modeling for Claim Severity Paper in Spring 2005 CAS Forum Glenn Meyers ISO Innovative Analytics Predictive Modeling Seminar September 19, 2005

Problems with Experience Rating for Excess of Loss Reinsurance Use submission claim severity data –Relevant, but –Not credible –Not developed Use industry distributions –Credible, but –Not relevant (???)

General Problems with Fitting Claim Severity Distributions Parameter uncertainty –Fitted parameters of chosen model are estimates subject to sampling error. Model uncertainty –We might choose the wrong model. There is no particular reason that the models we choose are appropriate. Loss development –Complete claim settlement data is not always available.

Outline of Talk Quantifying Parameter Uncertainty –Likelihood ratio test Incorporating Model Uncertainty –Use Bayesian estimation with likelihood functions –Uncertainty in excess layer loss estimates Bayesian estimation with prior models based on data reported to a statistical agent –Reflect insurer heterogeneity –Develops losses

The Likelihood Ratio Test

An Example – The Pareto Distribution Simulate random sample of size 1000  = 2.000,  = 10,000

Hypothesis Testing Example Significance level = 5%  2 critical value = H 0 : (  ) = (10000, 2) H 1 : (  ) ≠ (10000, 2) lnLR = 2( ) =1.207 Accept H 0

Hypothesis Testing Example Significance level = 5%  2 critical value = H 0 : (  ) = (10000, 1.7) H 1 : (  ) ≠ (10000, 1.7) lnLR = 2( ) = Reject H 0

Confidence Region X% confidence region corresponds to the 1-X% level hypothesis test. The set of all parameters (  ) that fail to reject corresponding H 0. For the 95% confidence region: –(10000, 2.0) is in. –(10000, 1.7) out.

Confidence Region Outer Ring 95%, Inner Ring 50%

Grouped Data Data grouped into four intervals –562 under 5000 –181 between 5000 and –134 between and –123 over Same data as before, only less information is given.

Confidence Region for Grouped Data Outer Ring 95%, Inner Ring 50%

Confidence Region for Ungrouped Data Outer Ring 95%, Inner Ring 50%

Estimation with Model Uncertainty COTOR Challenge – November 2004 COTOR published 250 claims –Distributional form not revealed to participants Participants were challenged to estimate the cost of a $5M x $5M layer. Estimate confidence interval for pure premium

You want to fit a distribution to 250 Claims Knee jerk first reaction, plot a histogram.

This will not do! Take logs And fit some standard distributions.

Still looks skewed. Take double logs. And fit some standard distributions.

Still looks skewed. Take triple logs. Still some skewness. Lognormal and gamma fits look somewhat better.

Candidate #1 Quadruple lognormal

Candidate #2 Triple loggamma

Candidate #3 Triple lognormal

All three cdf’s are within confidence interval for the quadruple lognormal.

Elements of Solution Three candidate models –Quadruple lognormal –Triple loggamma –Triple lognormal Parameter uncertainty within each model Construct a series of models consisting of –One of the three models. –Parameters within a broad confidence interval for each model. –7803 possible models

Steps in Solution Calculate likelihood (given the data) for each model. Use Bayes’ Theorem to calculate posterior probability for each model –Each model has equal prior probability.

Steps in Solution Calculate layer pure premium for 5 x 5 layer for each model. Expected pure premium is the posterior probability weighted average of the model layer pure premiums. Second moment of pure premium is the posterior probability weighted average of the model layer pure premiums squared.

CDF of Layer Pure Premium Probability that layer pure premium ≤ x equals Sum of posterior probabilities for which the model layer pure premium is ≤ x

Numerical Results

Histogram of Predictive Pure Premium

Example with Insurance Data Continue with Bayesian Estimation Liability insurance claim severity data Prior distributions derived from models based on individual insurer data Prior models reflect the maturity of claim data used in the estimation

Initial Insurer Models Selected 20 insurers –Claim count in the thousands Fit mixed exponential distribution to the data of each insurer Initial fits had volatile tails Truncation issues –Do small claims predict likelihood of large claims?

Initial Insurer Models

Low Truncation Point

High Truncation Point

Selections Made Truncation point = $100,000 Family of cdf’s that has “correct” behavior –Admittedly the definition of “correct” is debatable, but –The choices are transparent!

Selected Insurer Models

Each model consists of 1.The claim severity distribution for all claims settled within 1 year 2.The claim severity distribution for all claims settled within 2 years 3.The claim severity distribution for all claims settled within 3 years 4.The ultimate claim severity distribution for all claims 5.The ultimate limited average severity curve

Three Sample Insurers Small, Medium and Large Each has three years of data Calculate likelihood functions –Most recent year with #1 on prior slide –2 nd most recent year with #2 on prior slide –3 rd most recent year with #3 on prior slide Use Bayes theorem to calculate posterior probability of each model

Formulas for Posterior Probabilities Model (m) Cell Probabilities Likelihood (m) Using Bayes’ Theorem Number of claims

Results Taken from paper.

Formulas for Ultimate Layer Pure Premium Use #5 on model (3 rd previous) slide to calculate ultimate layer pure premium

Results All insurers were simulated from same population. Posterior standard deviation decreases with insurer size.

Possible Extensions Obtain model for individual insurers Obtain data for insurer of interest Calculate likelihood, Pr{data|model}, for each insurer’s model. Use Bayes’ Theorem to calculate posterior probability of each model Calculate the statistic of choice using models and posterior probabilities –e.g. Loss reserves