NAG Library Routine Document

e04ycf  (lsq_uncon_covariance)

 Contents

    1  Purpose
    7  Accuracy

1
Purpose

e04ycf returns estimates of elements of the variance-covariance matrix of the estimated regression coefficients for a nonlinear least squares problem. The estimates are derived from the Jacobian of the function fx at the solution.
This routine may be used following any one of the nonlinear least squares routines e04fcf, e04fyf, e04gbf, e04gdf, e04gyf, e04gzf, e04hef or e04hyf.

2
Specification

Fortran Interface
Subroutine e04ycf ( job, m, n, fsumsq, s, v, ldv, cj, work, ifail)
Integer, Intent (In):: job, m, n, ldv
Integer, Intent (Inout):: ifail
Real (Kind=nag_wp), Intent (In):: fsumsq, s(n)
Real (Kind=nag_wp), Intent (Inout):: v(ldv,n)
Real (Kind=nag_wp), Intent (Out):: cj(n), work(n)
C Header Interface
#include nagmk26.h
void  e04ycf_ ( const Integer *job, const Integer *m, const Integer *n, const double *fsumsq, const double s[], double v[], const Integer *ldv, double cj[], double work[], Integer *ifail)

3
Description

e04ycf is intended for use when the nonlinear least squares function, Fx=fTxfx, represents the goodness-of-fit of a nonlinear model to observed data. The routine assumes that the Hessian of Fx, at the solution, can be adequately approximated by 2JTJ, where J is the Jacobian of fx at the solution. The estimated variance-covariance matrix C is then given by
C=σ2JTJ-1,  JTJ  nonsingular,  
where σ2 is the estimated variance of the residual at the solution, x-, given by
σ2=Fx- m-n ,  
m being the number of observations and n the number of variables.
The diagonal elements of C are estimates of the variances of the estimated regression coefficients. See the E04 Chapter Introduction, Bard (1974) and Wolberg (1967) for further information on the use of C.
When JTJ is singular then C is taken to be
C=σ2 JTJ ,  
where JTJ  is the pseudo-inverse of JTJ, and
σ2 = F x- m-k ,   k=rank ​J  
but in this case the argument ifail is returned as nonzero as a warning to you that J has linear dependencies in its columns. The assumed rank of J can be obtained from ifail.
The routine can be used to find either the diagonal elements of C, or the elements of the jth column of C, or the whole of C.
e04ycf must be preceded by one of the nonlinear least squares routines mentioned in Section 1, and requires the arguments fsumsq, s and v to be supplied by those routines (e.g., see e04fcf). fsumsq is the residual sum of squares Fx- and s and v contain the singular values and right singular vectors respectively in the singular value decomposition of J. s and v are returned directly by the comprehensive routines e04fcf, e04gbf, e04gdf and e04hef, but are returned as part of the workspace argument w (from one of the easy-to-use routines e04fyf, e04gyf, e04gzf and e04hyf). In the case of e04fyf, s starts at wNS, where
NS=6× n+2× m+m× n+1+ max1, n × n-1 / 2  
and in the cases of the remaining easy-to-use routines, s starts at wNS, where
NS=7× n+2× m+m× n+n× n+1 / 2+1+ max1, n × n-1 / 2 .  
The argument v starts immediately following the elements of s, so that v starts at wNV, where
NV=NS+n .  
For all the easy-to-use routines the argument ldv must be supplied as n. Thus a call to e04ycf following e04fyf can be illustrated as
.
.
.
Call e04fyf (m, n, lfun1, x, fsumsq, w, lw, iuser, ruser, ifail)
.
.
.
ns = 6*n + 2*m + m*n + 1 + max(1,n*(n-1)/2)
nv = ns + n
Call e04ycf (job, m, n, fsumsq, w(ns), w(nv), n, cj, work, ifail)
where the arguments m, n, fsumsq and the n+n2 elements wNS , wNS+1 , , w NV+ n2 -1  must not be altered between the calls to e04fyf and e04ycf. The above illustration also holds for a call to e04ycf following a call to one of e04gyf, e04gzf or e04hyf, except that NS must be computed as
NS=7× n+2× m+m× n+ n× n+1 /2+1+ max 1, n × n-1 / 2 .  

4
References

Bard Y (1974) Nonlinear Parameter Estimation Academic Press
Wolberg J R (1967) Prediction Analysis Van Nostrand

5
Arguments

1:     job – IntegerInput
On entry: which elements of C are returned as follows:
job=-1
The n by n symmetric matrix C is returned.
job=0
The diagonal elements of C are returned.
job>0
The elements of column job of C are returned.
Constraint: -1jobn.
2:     m – IntegerInput
On entry: the number m of observations (residuals fix).
Constraint: mn.
3:     n – IntegerInput
On entry: the number n of variables xj.
Constraint: 1nm.
4:     fsumsq – Real (Kind=nag_wp)Input
On entry: the sum of squares of the residuals, Fx-, at the solution x-, as returned by the nonlinear least squares routine.
Constraint: fsumsq0.0.
5:     sn – Real (Kind=nag_wp) arrayInput
On entry: the n singular values of the Jacobian as returned by the nonlinear least squares routine. See Section 3 for information on supplying s following one of the easy-to-use routines.
6:     vldvn – Real (Kind=nag_wp) arrayInput/Output
On entry: the n by n right-hand orthogonal matrix (the right singular vectors) of J as returned by the nonlinear least squares routine. See Section 3 for information on supplying v following one of the easy-to-use routines.
On exit: if job0, v is unchanged.
If job=-1, the leading n by n part of v is overwritten by the n by n matrix C. When e04ycf is called with job=-1 following an easy-to-use routine this means that C is returned, column by column, in the n2 elements of w given by wNV,wNV+1,,wNV+n2-1. (See Section 3 for the definition of NV.)
7:     ldv – IntegerInput
On entry: the first dimension of the array v as declared in the (sub)program from which e04ycf is called. When v is passed in the workspace argument w (following one of the easy-to-use least square routines), ldv must be the value n.
Constraint: ldvn.
8:     cjn – Real (Kind=nag_wp) arrayOutput
On exit: if job=0, cj returns the n diagonal elements of C.
If job=j>0, cj returns the n elements of the jth column of C.
If job=-1, cj is not referenced.
9:     workn – Real (Kind=nag_wp) arrayWorkspace
If job=-1 or 0, work is used as internal workspace.
If job>0, work is not referenced.
10:   ifail – IntegerInput/Output
On entry: ifail must be set to 0, -1​ or ​1. If you are unfamiliar with this argument you should refer to Section 3.4 in How to Use the NAG Library and its Documentation for details.
For environments where it might be inappropriate to halt program execution when an error is detected, the value -1​ or ​1 is recommended. If the output of error messages is undesirable, then the value 1 is recommended. Otherwise, because for this routine the values of the output arguments may be useful even if ifail0 on exit, the recommended value is -1. When the value -1​ or ​1 is used it is essential to test the value of ifail on exit.
On exit: ifail=0 unless the routine detects an error or a warning has been flagged (see Section 6).

6
Error Indicators and Warnings

If on entry ifail=0 or -1, explanatory error messages are output on the current error message unit (as defined by x04aaf).
Note: e04ycf may return useful information for one or more of the following detected errors or warnings.
Errors or warnings detected by the routine:
ifail=1
On entry,job<-1,
orjob>n,
orn<1,
orm<n,
orfsumsq<0.0,
orldv<n.
ifail=2
The singular values are all zero, so that at the solution the Jacobian matrix J has rank 0.
ifail>2
At the solution the Jacobian matrix contains linear, or near linear, dependencies amongst its columns. In this case the required elements of C have still been computed based upon J having an assumed rank given by ifail-2. The rank is computed by regarding singular values SVj that are not larger than 10ε×SV1 as zero, where ε is the machine precision (see x02ajf). If you expect near linear dependencies at the solution and are happy with this tolerance in determining rank you should call e04ycf with ifail=1 in order to prevent termination (see the description of ifail). It is then essential to test the value of ifail on exit from e04ycf.
Overflow
If overflow occurs then either an element of C is very large, or the singular values or singular vectors have been incorrectly supplied.
ifail=-99
An unexpected error has been triggered by this routine. Please contact NAG.
See Section 3.9 in How to Use the NAG Library and its Documentation for further information.
ifail=-399
Your licence key may have expired or may not have been installed correctly.
See Section 3.8 in How to Use the NAG Library and its Documentation for further information.
ifail=-999
Dynamic memory allocation failed.
See Section 3.7 in How to Use the NAG Library and its Documentation for further information.

7
Accuracy

The computed elements of C will be the exact covariances corresponding to a closely neighbouring Jacobian matrix J.

8
Parallelism and Performance

e04ycf is threaded by NAG for parallel execution in multithreaded implementations of the NAG Library.
e04ycf makes calls to BLAS and/or LAPACK routines, which may be threaded within the vendor library used by this implementation. Consult the documentation for the vendor library for further information.
Please consult the X06 Chapter Introduction for information on how to control and interrogate the OpenMP environment used within this routine. Please also consult the Users' Note for your implementation for any additional implementation-specific information.

9
Further Comments

When job=-1 the time taken by e04ycf is approximately proportional to n3. When job0 the time taken by the routine is approximately proportional to n2.

10
Example

This example estimates the variance-covariance matrix C for the least squares estimates of x1, x2 and x3 in the model
y=x1+t1x2t2+x3t3  
using the 15 sets of data given in the following table:
y t1 t2 t3 0.14 1.0 15.0 1.0 0.18 2.0 14.0 2.0 0.22 3.0 13.0 3.0 0.25 4.0 12.0 4.0 0.29 5.0 11.0 5.0 0.32 6.0 10.0 6.0 0.35 7.0 9.0 7.0 0.39 8.0 8.0 8.0 0.37 9.0 7.0 7.0 0.58 10.0 6.0 6.0 0.73 11.0 5.0 5.0 0.96 12.0 4.0 4.0 1.34 13.0 3.0 3.0 2.10 14.0 2.0 2.0 4.39 15.0 1.0 1.0  
The program uses 0.5,1.0,1.5 as the initial guess at the position of the minimum and computes the least squares solution using e04fyf. See the routine document e04fyf for further information.

10.1
Program Text

Program Text (e04ycfe.f90)

10.2
Program Data

Program Data (e04ycfe.d)

10.3
Program Results

Program Results (e04ycfe.r)

© The Numerical Algorithms Group Ltd, Oxford, UK. 2017