nag_censored_normal (g07bbc) (PDF version)
g07 Chapter Contents
g07 Chapter Introduction
NAG C Library Manual

NAG Library Function Document

nag_censored_normal (g07bbc)

+ Contents

    1  Purpose
    7  Accuracy

1  Purpose

nag_censored_normal (g07bbc) computes maximum likelihood estimates and their standard errors for arguments of the Normal distribution from grouped and/or censored data.

2  Specification

#include <nag.h>
#include <nagg07.h>
void  nag_censored_normal (Nag_CEMethod method, Integer n, const double x[], const double xc[], const Integer ic[], double *xmu, double *xsig, double tol, Integer maxit, double *sexmu, double *sexsig, double *corr, double *dev, Integer nobs[], Integer *nit, NagError *fail)

3  Description

A sample of size n is taken from a Normal distribution with mean μ and variance σ2 and consists of grouped and/or censored data. Each of the n observations is known by a pair of values Li,Ui such that:
LixiUi.
The data is represented as particular cases of this form:
Let the set A identify the exactly specified observations, sets B and C identify the observations censored on the right and left respectively, and set D identify the observations confined between two finite limits. Also let there be r exactly specified observations, i.e., the number in A. The probability density function for the standard Normal distribution is
Zx=12πexp-12x2 ,  -<x<
and the cumulative distribution function is
PX= 1-QX=-XZxdx.
The log-likelihood of the sample can be written as:
L μ,σ =-r logσ - 1 2 A x i -μ / σ 2 +B log Q l i + C log P u i + D log p i
where pi=Pui-Pli and ui=Ui-μ/σ,  li=Li-μ/σ.
Let
Sxi=Zxi Qxi ,  S1li,ui=Zli-Zuipi
and
S2li,ui=uiZui-liZlipi,
then the first derivatives of the log-likelihood can be written as:
Lμ,σ μ =L1μ,σ=σ-2Axi-μ+σ-1BSli-σ-1CS-ui+σ-1DS1li,ui
and
Lμ,σ σ =L2μ,σ=-rσ-1+σ-3A xi-μ 2+σ-1BliSli-σ-1CuiS-ui
-σ-1DS2li,ui
The maximum likelihood estimates, μ^ and σ^, are the solution to the equations:
L1μ^,σ^=0 (1)
and
L2μ^,σ^=0 (2)
and if the second derivatives 2 L 2 μ , 2 L μ σ  and 2L 2σ  are denoted by L11, L12 and L22 respectively, then estimates of the standard errors of μ^ and σ^ are given by:
seμ^=-L22 L11L22-L122 ,  seσ^=-L11 L11L22-L122
and an estimate of the correlation of μ^ and σ^ is given by:
L12L12L22 .
To obtain the maximum likelihood estimates the equations (1) and (2) can be solved using either the Newton–Raphson method or the Expectation-maximization EM algorithm of Dempster et al. (1977).
Newton–Raphson Method
This consists of using approximate estimates μ~ and σ~ to obtain improved estimates μ~+δμ~ and σ~+δσ~ by solving
δμ~L11+δσ~L12+L1=0, δμ~L12+δσ~L22+L2=0,
for the corrections δμ~ and δσ~.
EM Algorithm
The expectation step consists of constructing the variable wi as follows:
if   iA,   wi= xi E Li< xi< Ui = μ+σ S1 li,ui (3)
if   iB,   wi= E xi xi> Li =μ+σS li S1 li,ui (4)
if   iC,   wi= E xi xi< Ui =μ-σS -ui li,ui (5)
if   iD,   wi= E xi Li< xi< Ui =μ+σ S1 li,ui (6)
the maximization step consists of substituting (3), (4), (5) and (6) into (1) and (2) giving:
μ^=i=1nw^i/n (7)
and
σ^2=i=1nw^i-μ^2/ r+BTl^i+CT-u^i+DT1l^i,u^i (8)
where
Tx=SxSx-x ,   T1l,u=S12l,u+S2l,u
and where w^i, l^i and u^i are wi, li and ui evaluated at μ^ and σ^. Equations (3) to (8) are the basis of the EM iterative procedure for finding μ^ and σ^2. The procedure consists of alternately estimating μ^ and σ^2 using (7) and (8) and estimating w^i using (3) to (6).
In choosing between the two methods a general rule is that the Newton–Raphson method converges more quickly but requires good initial estimates whereas the EM algorithm converges slowly but is robust to the initial values. In the case of the censored Normal distribution, if only a small proportion of the observations are censored then estimates based on the exact observations should give good enough initial estimates for the Newton–Raphson method to be used. If there are a high proportion of censored observations then the EM algorithm should be used and if high accuracy is required the subsequent use of the Newton–Raphson method to refine the estimates obtained from the EM algorithm should be considered.

4  References

Dempster A P, Laird N M and Rubin D B (1977) Maximum likelihood from incomplete data via the EM algorithm (with discussion) J. Roy. Statist. Soc. Ser. B 39 1–38
Swan A V (1969) Algorithm AS 16. Maximum likelihood estimation from grouped and censored normal data Appl. Statist. 18 110–114
Wolynetz M S (1979) Maximum likelihood estimation from confined and censored normal data Appl. Statist. 28 185–195

5  Arguments

1:     methodNag_CEMethodInput
On entry: indicates whether the Newton–Raphson or EM algorithm should be used.
If method=Nag_CE_NR, then the Newton–Raphson algorithm is used.
If method=Nag_CE_EM, then the EM algorithm is used.
Constraint: method=Nag_CE_NR or Nag_CE_EM.
2:     nIntegerInput
On entry: n, the number of observations.
Constraint: n2.
3:     x[n]const doubleInput
On entry: the observations xi, Li or Ui, for i=1,2,,n.
If the observation is exactly specified – the exact value, xi.
If the observation is right-censored – the lower value, Li.
If the observation is left-censored – the upper value, Ui.
If the observation is interval-censored – the lower or upper value, Li or Ui, (see xc).
4:     xc[n]const doubleInput
On entry: if the jth observation, for j=1,2,,n is an interval-censored observation then xc[j-1] should contain the complementary value to x[j-1], that is, if x[j-1]<xc[j-1], then xc[j-1] contains upper value, Ui, and if x[j-1]>xc[j-1], then xc[j-1] contains lower value, Li. Otherwise if the jth observation is exact or right- or left-censored xc[j-1] need not be set.
Note: if x[j-1]=xc[j-1] then the observation is ignored.
5:     ic[n]const IntegerInput
On entry: ic[i-1] contains the censoring codes for the ith observation, for i=1,2,,n.
If ic[i-1]=0, the observation is exactly specified.
If ic[i-1]=1, the observation is right-censored.
If ic[i-1]=2, the observation is left-censored.
If ic[i-1]=3, the observation is interval-censored.
Constraint: ic[i-1]=0, 1, 2 or 3, for i=1,2,,n.
6:     xmudouble *Input/Output
On entry: if xsig>0.0 the initial estimate of the mean, μ; otherwise xmu need not be set.
On exit: the maximum likelihood estimate, μ^, of μ.
7:     xsigdouble *Input/Output
On entry: specifies whether an initial estimate of μ and σ are to be supplied.
xsig>0.0
xsig is the initial estimate of σ and xmu must contain an initial estimate of μ.
xsig0.0
Initial estimates of xmu and xsig are calculated internally from:
(a) the exact observations, if the number of exactly specified observations is 2; or
(b) the interval-censored observations; if the number of interval-censored observations is 1; or
(c) they are set to 0.0 and 1.0 respectively.
On exit: the maximum likelihood estimate, σ^, of σ.
8:     toldoubleInput
On entry: the relative precision required for the final estimates of μ and σ. Convergence is assumed when the absolute relative changes in the estimates of both μ and σ are less than tol.
If tol=0.0, then a relative precision of 0.000005 is used.
Constraint: machine precision<tol1.0 or tol=0.0.
9:     maxitIntegerInput
On entry: the maximum number of iterations.
If maxit0, then a value of 25 is used.
10:   sexmudouble *Output
On exit: the estimate of the standard error of μ^.
11:   sexsigdouble *Output
On exit: the estimate of the standard error of σ^.
12:   corrdouble *Output
On exit: the estimate of the correlation between μ^ and σ^.
13:   devdouble *Output
On exit: the maximized log-likelihood, Lμ^,σ^.
14:   nobs[4]IntegerOutput
On exit: the number of the different types of each observation;
nobs[0] contains number of right-censored observations.
nobs[1] contains number of left-censored observations.
nobs[2] contains number of interval-censored observations.
nobs[3] contains number of exactly specified observations.
15:   nitInteger *Output
On exit: the number of iterations performed.
16:   failNagError *Input/Output
The NAG error argument (see Section 3.6 in the Essential Introduction).

6  Error Indicators and Warnings

NE_ALLOC_FAIL
Dynamic memory allocation failed.
NE_BAD_PARAM
On entry, argument value had an illegal value.
NE_CONVERGENCE
Method has not converged in value iterations.
NE_DIVERGENCE
Process has diverged.
NE_EM_PROCESS
The EM process has failed.
NE_INT
On entry, n=value.
Constraint: n2.
NE_INT_ARRAY
On entry, ic[value] is invalid, it contains value.
NE_INTERNAL_ERROR
An internal error has occurred in this function. Check the function call and any array sizes. If the call is correct then please contact NAG for assistance.
NE_OBSERVATIONS
On entry, effective number of observations <2.
NE_REAL
On entry, tol is invalid: tol=value.
NE_STANDARD_ERRORS
Standard errors cannot be computed.

7  Accuracy

The accuracy is controlled by the argument tol.
If high precision is requested with the EM algorithm then there is a possibility that, due to the slow convergence, before the correct solution has been reached the increments of μ^ and σ^ may be smaller than tol and the process will prematurely assume convergence.

8  Further Comments

The process is deemed divergent if three successive increments of μ or σ increase.

9  Example

A sample of 18 observations and their censoring codes are read in and the Newton–Raphson method used to compute the estimates.

9.1  Program Text

Program Text (g07bbce.c)

9.2  Program Data

Program Data (g07bbce.d)

9.3  Program Results

Program Results (g07bbce.r)


nag_censored_normal (g07bbc) (PDF version)
g07 Chapter Contents
g07 Chapter Introduction
NAG C Library Manual

© The Numerical Algorithms Group Ltd, Oxford, UK. 2012