Scribd
Upload
12 views11 pages

Parametric Models in Statistics Overview

Parametric models assume that data belongs to a certain distribution family like normal or exponential. They are easy to use but have errors from incorrect parameters or models. Non-parametric models make no assumptions and adhere closely to data, but are hard to make inferences from. Examples of parametric survival analysis include the uniform, exponential, Gompertz and Weibull distributions, while the Lee-Carter model accounts for cohort effects over time. Parametric models with explanatory factors include linear log-time and accelerated failure time models.

Uploaded by

Kenneth Wong
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
12 views11 pages

Parametric Models in Statistics Overview

Parametric models assume that data belongs to a certain distribution family like normal or exponential. They are easy to use but have errors from incorrect parameters or models. Non-parametric models make no assumptions and adhere closely to data, but are hard to make inferences from. Examples of parametric survival analysis include the uniform, exponential, Gompertz and Weibull distributions, while the Lee-Carter model accounts for cohort effects over time. Parametric models with explanatory factors include linear log-time and accelerated failure time models.

Uploaded by

Kenneth Wong
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd

Overview of parametric models

Overview of parametric models:


Introduction


c University of New South Wales
School of Risk and Actuarial Studies

1/11
Overview of parametric models
Introduction

Parametric models

assume that the distribution belongs to a certain family (e.g.


normal or exponential)
advantages
generally easy interpolation/extrapolation
generally easy to draw statistical inference
easy way of smoothing the data
disadvantages
parameter error
model error

2/11
Overview of parametric models
Introduction

Non-parametric models

Non-parametric model
no prior assumptions about the shape or form of the
distribution are made
advantage
adherence to data
disadvantages
may be hard to make inference
need extra assumptions to interpolate/extrapolate the data

3/11
Overview of parametric models
Introduction

Some examples

Non-parametric
estimating a probability density/mass function with a
histogram
modelling a survival function by using the Kaplan-Meier
method
Semi-parametric
modelling the mortality rate with a Cox regression model
Parametric
fitting data with a parametric distribution (e.g. Gamma,
Weibull, copula, etc.)
parametric regression models

4/11
Overview of parametric models
Introduction

Survival analysis with parametric distributions


Uniform distribution (de Moivre)

1, x 0

S(x; ) = x , 0<x

0, x >

1
x =
x
Exponential distribution (constant mortality rate/memoryless)

S(x; ) = e x , t > 0

x =
5/11
Overview of parametric models
Introduction

Survival analysis with parametric distributions

type I extreme least value

 
R ax
S(x; a, R) = exp e , < x < , a > 0, R > 0
a

x = Re ax

6/11
Overview of parametric models
Introduction

Survival analysis with parametric distributions

Gompertz distribution
a truncated (from 0) modification of the type I extreme least
value distribution
 
R ax
S(x; a, R) = exp (1 e ) , x > 0
a

x = Re ax
Makeham-Gompertz distribution
allowing for accidental deaths in addition to deaths from
natural causes
x = A + Re ax

7/11
Overview of parametric models
Introduction

Survival analysis with parametric distributions

Other examples include


modelling a life time distribution: Weibull, log-Normal,
Gamma, logistic
modelling stochastic transitions of states: Markov chain

8/11
Overview of parametric models
Introduction

Survival analysis with cohort effects

Lee-Carter model

x (t) = x + x t + x,t

where
x represents the age-pattern
x represents the deviations from the averaged pattern
t represents the change in level of mortality over time
x,t represents the error term of age-specific fluctuation

9/11
Overview of parametric models
Introduction

Survival analysis with explanatory factors

Linear log-time model


multiplicative effects on the survival time (via additive effects
on the logarithm of the survival time)
Accelerated failure time model
generalisation of the linear log-time model
covariates scales (in some functional way) the survival time
(given a baseline model)

10/11
Overview of parametric models
Introduction

Additional reference

Elandt-Johnson, R.C. and Johnson, N.L.. Survival models and data


analysis, Wiley & Sons, Inc., New York, 1999.
Chapter 3: Survival distributions.

11/11

You might also like