Scribd
Upload
49 views

PRINCIPLES OF LEAST SQUARES Lecture Notes

Teacher format

Uploaded by

wheelergeorge942
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
49 views

PRINCIPLES OF LEAST SQUARES Lecture Notes

Teacher format

Uploaded by

wheelergeorge942
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 3

Generalized Least Squares (GLS)

Generalized Least Squares (GLS) is a statistical method used to estimate the unknown parameters in

a linear regression model. It’s particularly useful when there is a correlation between the residuals

(errors) in the regression model, which violates the assumptions of Ordinary Least Squares (OLS)

regression.

Key Points of GLS:

1. Purpose: GLS is designed to handle situations where the residuals are heteroskedastic

(having different variances) or autocorrelated (correlated with each other). This makes GLS

more efficient and reliable than OLS in such cases1.

2. Model Setup: In a standard linear regression model, you have:

o Response variable: ( y )

o Predictor variables: ( X )

o Regression coefficients: ( \beta )

o Error term: ( \epsilon )

The model can be written as:

y = X\beta + \epsilon

3. Assumptions: GLS assumes that the error terms have a known covariance matrix ( \Sigma ).

If ( \Sigma ) is unknown, it can be estimated, leading to Feasible Generalized Least Squares

(FGLS)2.

4. Estimation: The GLS estimator minimizes the weighted sum of squared residuals, taking into

account the covariance structure of the errors. The estimator is given by:

\hat{\beta}_{GLS} = (X^T \Sigma^{-1} X)^{-1} X^T \Sigma^{-1} y

5. Advantages: GLS provides unbiased, consistent, and efficient estimates of the regression

coefficients, making it a powerful tool when dealing with non-constant error variances or

correlated errors2.

Would you like to know more about how GLS is applied in practice or its assumptions?
Solution of Normal Equations Treatment of large geodetic networks

The solution of normal equations in the treatment of large geodetic networks involves sophisticated

techniques to handle the vast amount of data and the complexity of the calculations. Here are some

key points:

Key Concepts:

1. Normal Equations: In the context of least squares adjustment, normal equations are derived

from the condition that the sum of the squared residuals (differences between observed and

computed values) is minimized. For a system of linear equations (Ax = b), the normal

equations are given by:

A^T A x = A^T b

2. Large Geodetic Networks: These involve a large number of observations and unknowns,

making the system of equations very large and sparse. Efficient algorithms are needed to

solve these systems.

Techniques for Solving Normal Equations:

1. Block-Orthogonal Decomposition: This method involves breaking down the large system into

smaller, more manageable blocks. It helps in efficient data management and numerical

stability1.

2. Nested Dissection: This is a technique used to reorder the equations to minimize fill-in (non-

zero elements) during the factorization process. It helps in reducing computational

complexity1.

3. Helmert Blocking: This method partitions the network into smaller blocks that can be solved

independently and then combined. It is particularly useful for very large-scale problems2.

4. Parallel Processing: Modern approaches often use parallel processing to handle the large

datasets efficiently. This involves distributing the computations across multiple processors3.

Practical Application:
 Example: The readjustment of the North American Datum in 1983 involved solving an

overdetermined system of approximately 1,000,000 equations in 400,000 unknowns.

Techniques like block-orthogonal decomposition and nested dissection were crucial in

managing this large-scale problem1.

Would you like more details on any specific technique or a practical example of how these methods

are applied?

1
: Stanford University 2: Tandfonline 3: MyGeodesy

You might also like