USGS

Regression Modeling of Ground-Water Flow

U.S. Geological Survey, Techniques of Water-Resources Investigations, Book 3, Chapter B4

By Richard L. Cooley and Richard L. Naff


Table of Contents

Introduction

Flow equation and boundary conditions

Types of solutions

Direct solution for head

Inverse solution for parameters

Solution using real data

Sources of error in ground-water data

Sources of error in head data

Sources of error in parameter data

Model construction

Trial and error methods

Formal optimization procedures

References cited

Additional reading

Review of probability and statistics

Basic concepts

Frequencies and distributions

Discrete random variables

Problem 2.2-1

Histograms

Continuous random variables

Problem 2.2-2

Properties of cumulative distribution functions

An example: the normal distribution

Expectation and the continuous random variable

The mean

Problem 2.3-l

Generalization and application of the expectation operator

The variance, standard deviation, and coefficient of variation

Problem 2.3-2

Jointly distributed random variables

Expectation of jointly distributed random variables

Independent random variables

Conditional probabilities

Problem 2.4-l

Variance of a column vector

Problem 2.4-2

Estimators of population parameters

Mean estimator

Problem 2.5-l

Variance estimator

Estimator of correlation coefficient

Summary

Problem 2.5-2

Transformation of random variables

Sum of independent normal random variables

The Chi-square distribution

The F distribution

Problem 2.6-l

Central limit theorem

Confidence limits

Problem 2.8-l

Hypothesis testing

Type I error

One-tailed test

Two-tailed test

Type II error

Summary of method

Problem 2.9-l

Tables of probability distributions

Appendices

Correlation of two linearly related random variables

Expected value of variance estimator

References cited

Additional reading

Regression solution of modeling problems

Introduction and background

Assumed model structure

Least-squares estimation

Inclusion of prior information

Problem 3.1-l

Regression when the model is linear

Derivation of solution

Solution algorithm

Problem 3.2-l

Singularity and conditioning

Regression when the model is nonlinear

Modified Gauss-Newton method

Problem 3.3-l

3.3.2 Non-linear regression when the model is numerical

Problem 3.3-2

Convergence and conditioning

Computation of and u and p

Regression including prior information

Model structure

Solution procedures

References cited

Additional reading

Numerical nonlinear regression solution of general steady-state ground-water flow problems

Assumed model and solution procedure

Problem specification

Matrix form of regression model

Nonlinear regression solution

Singularity and conditioning

Problem 4.2-l

Problem 4.2-2

Appendices

Integrated finite difference model

Computation of nodal sensitivities for the integrated finite difference model

Derivation of equation 4.2-l

Documentation of program for nonlinear regression solution of steady-state ground-water flow problems

References cited

Additional reading

Elementary analysis and use of the regression model

Assumed forms of model equations

Assumptions of regression modeling

Relationships between residuals and disturbances

Some statistical measures

The error variance

The correlation, Ry, between wY and wt

The variance-covariance matrix for b

The correlation, rij, between any two parameters bi and bj

Problem 5.4-l

Analysis of residuals

Distribution of residuals

Graphical procedures

Problem 5.5-l

Problem 5.5-2

Investigation of alternative parameter sets

Generalized W statistic

Joint confidence region for B2

Problem 5.6-l

Problem 5.6-2

Problem 5.6-3

Investigation of predictive reliability

The variance-covariance matrix for t

5.7.2 Confidence interval for fBi

5.7.3 Prediction interval for predicted observation Yjpred

Problem 5.7-l

Appendix

Documentation of program to compute vectors d and g of section 5.5.2

References cited

Additional reading

Some advanced topics

Advanced models

Regression when the dependent variable is implicit

Regression when the implicit-variable model is numerical

Modified Beale's measure of nonlinearity

Problem 6.2-l

Problem 6.2-2

Compatibility of prior and regression estimates of parameters

Problem 6.3-l

Appendix

Documentation of program to compute the modified Beale’s measure

References cited

Additional reading

Answers to exercises

 


Back




FirstGov button  Take Pride in America button