Linear least squares

(Redirected from Normal equation)

Linear least squares (LLS) is the least squares approximation of linear functions to data. It is a set of formulations for solving statistical problems involved in linear regression, including variants for ordinary (unweighted), weighted, and generalized (correlated) residuals. Numerical methods for linear least squares include inverting the matrix of the normal equations and orthogonal decomposition methods.

Basic formulation

edit

Consider the linear equation

  (1)

where   and   are given and   is variable to be computed. When   it is generally the case that (1) has no solution. For example, there is no value of   that satisfies   because the first two rows require that   but then the third row is not satisfied. Thus, for   the goal of solving (1) exactly is typically replaced by finding the value of   that minimizes some error. There are many ways that the error can be defined, but one of the most common is to define it as   This produces a minimization problem, called a least squares problem

  (2)

The solution to the least squares problem (1) is computed by solving the normal equation[1]

  (3)

where   denotes the transpose of  .

Continuing the example, above, with   we find   and   Solving the normal equation gives  

Formulations for Linear Regression

edit

The three main linear least squares formulations are:

  • Ordinary least squares (OLS) is the most common estimator. OLS estimates are commonly used to analyze both experimental and observational data.
    The OLS method minimizes the sum of squared residuals, and leads to a closed-form expression for the estimated value of the unknown parameter vector β:   where   is a vector whose ith element is the ith observation of the dependent variable, and   is a matrix whose ij element is the ith observation of the jth independent variable. The estimator is unbiased and consistent if the errors have finite variance and are uncorrelated with the regressors:[2]   where   is the transpose of row i of the matrix   It is also efficient under the assumption that the errors have finite variance and are homoscedastic, meaning that E[εi2|xi] does not depend on i. The condition that the errors are uncorrelated with the regressors will generally be satisfied in an experiment, but in the case of observational data, it is difficult to exclude the possibility of an omitted covariate z that is related to both the observed covariates and the response variable. The existence of such a covariate will generally lead to a correlation between the regressors and the response variable, and hence to an inconsistent estimator of β. The condition of homoscedasticity can fail with either experimental or observational data. If the goal is either inference or predictive modeling, the performance of OLS estimates can be poor if multicollinearity is present, unless the sample size is large.
  • Weighted least squares (WLS) are used when heteroscedasticity is present in the error terms of the model.
  • Generalized least squares (GLS) is an extension of the OLS method, that allows efficient estimation of β when either heteroscedasticity, or correlations, or both are present among the error terms of the model, as long as the form of heteroscedasticity and correlation is known independently of the data. To handle heteroscedasticity when the error terms are uncorrelated with each other, GLS minimizes a weighted analogue to the sum of squared residuals from OLS regression, where the weight for the ith case is inversely proportional to var(εi). This special case of GLS is called "weighted least squares". The GLS solution to an estimation problem is   where Ω is the covariance matrix of the errors. GLS can be viewed as applying a linear transformation to the data so that the assumptions of OLS are met for the transformed data. For GLS to be applied, the covariance structure of the errors must be known up to a multiplicative constant.

Alternative formulations

edit

Other formulations include:

  • Iteratively reweighted least squares (IRLS) is used when heteroscedasticity, or correlations, or both are present among the error terms of the model, but where little is known about the covariance structure of the errors independently of the data.[3] In the first iteration, OLS, or GLS with a provisional covariance structure is carried out, and the residuals are obtained from the fit. Based on the residuals, an improved estimate of the covariance structure of the errors can usually be obtained. A subsequent GLS iteration is then performed using this estimate of the error structure to define the weights. The process can be iterated to convergence, but in many cases, only one iteration is sufficient to achieve an efficient estimate of β.[4][5]
  • Instrumental variables regression (IV) can be performed when the regressors are correlated with the errors. In this case, we need the existence of some auxiliary instrumental variables zi such that E[ziεi] = 0. If Z is the matrix of instruments, then the estimator can be given in closed form as   Optimal instruments regression is an extension of classical IV regression to the situation where E[εi | zi] = 0.
  • Total least squares (TLS)[6] is an approach to least squares estimation of the linear regression model that treats the covariates and response variable in a more geometrically symmetric manner than OLS. It is one approach to handling the "errors in variables" problem, and is also sometimes used even when the covariates are assumed to be error-free.
  • Linear Template Fit (LTF)[7] combines a linear regression with (generalized) least squares in order to determine the best estimator. The Linear Template Fit addresses the frequent issue, when the residuals cannot be expressed analytically or are too time consuming to be evaluate repeatedly, as it is often the case in iterative minimization algorithms. In the Linear Template Fit, the residuals are estimated from the random variables and from a linear approximation of the underlying true model, while the true model needs to be provided for at least   (were   is the number of estimators) distinct reference values β. The true distribution is then approximated by a linear regression, and the best estimators are obtained in closed form as   where   denotes the template matrix with the values of the known or previously determined model for any of the reference values β,   are the random variables (e.g. a measurement), and the matrix   and the vector   are calculated from the values of β. The LTF can also be expressed for Log-normal distribution distributed random variables. A generalization of the LTF is the Quadratic Template Fit, which assumes a second order regression of the model, requires predictions for at least   distinct values β, and it finds the best estimator using Newton's method.
  • Percentage least squares focuses on reducing percentage errors, which is useful in the field of forecasting or time series analysis. It is also useful in situations where the dependent variable has a wide range without constant variance, as here the larger residuals at the upper end of the range would dominate if OLS were used. When the percentage or relative error is normally distributed, least squares percentage regression provides maximum likelihood estimates. Percentage regression is linked to a multiplicative error model, whereas OLS is linked to models containing an additive error term.[8]

Objective function

edit

In OLS (i.e., assuming unweighted observations), the optimal value of the objective function is found by substituting the optimal expression for the coefficient vector:   where  , the latter equality holding since   is symmetric and idempotent. It can be shown from this[9] that under an appropriate assignment of weights the expected value of S is  . If instead unit weights are assumed, the expected value of S is  , where   is the variance of each observation.

If it is assumed that the residuals belong to a normal distribution, the objective function, being a sum of weighted squared residuals, will belong to a chi-squared ( ) distribution with m − n degrees of freedom. Some illustrative percentile values of   are given in the following table.[10]

       
10 9.34 18.3 23.2
25 24.3 37.7 44.3
100 99.3 124 136

These values can be used for a statistical criterion as to the goodness of fit. When unit weights are used, the numbers should be divided by the variance of an observation.

For WLS, the ordinary objective function above is replaced for a weighted average of residuals.

Discussion

edit

In statistics and mathematics, linear least squares is an approach to fitting a mathematical or statistical model to data in cases where the idealized value provided by the model for any data point is expressed linearly in terms of the unknown parameters of the model. The resulting fitted model can be used to summarize the data, to predict unobserved values from the same system, and to understand the mechanisms that may underlie the system.

Mathematically, linear least squares is the problem of approximately solving an overdetermined system of linear equations A x = b, where b is not an element of the column space of the matrix A. The approximate solution is realized as an exact solution to A x = b', where b' is the projection of b onto the column space of A. The best approximation is then that which minimizes the sum of squared differences between the data values and their corresponding modeled values. The approach is called linear least squares since the assumed function is linear in the parameters to be estimated. Linear least squares problems are convex and have a closed-form solution that is unique, provided that the number of data points used for fitting equals or exceeds the number of unknown parameters, except in special degenerate situations. In contrast, non-linear least squares problems generally must be solved by an iterative procedure, and the problems can be non-convex with multiple optima for the objective function. If prior distributions are available, then even an underdetermined system can be solved using the Bayesian MMSE estimator.

In statistics, linear least squares problems correspond to a particularly important type of statistical model called linear regression which arises as a particular form of regression analysis. One basic form of such a model is an ordinary least squares model. The present article concentrates on the mathematical aspects of linear least squares problems, with discussion of the formulation and interpretation of statistical regression models and statistical inferences related to these being dealt with in the articles just mentioned. See outline of regression analysis for an outline of the topic.

Properties

edit

If the experimental errors,  , are uncorrelated, have a mean of zero and a constant variance,  , the Gauss–Markov theorem states that the least-squares estimator,  , has the minimum variance of all estimators that are linear combinations of the observations. In this sense it is the best, or optimal, estimator of the parameters. Note particularly that this property is independent of the statistical distribution function of the errors. In other words, the distribution function of the errors need not be a normal distribution. However, for some probability distributions, there is no guarantee that the least-squares solution is even possible given the observations; still, in such cases it is the best estimator that is both linear and unbiased.

For example, it is easy to show that the arithmetic mean of a set of measurements of a quantity is the least-squares estimator of the value of that quantity. If the conditions of the Gauss–Markov theorem apply, the arithmetic mean is optimal, whatever the distribution of errors of the measurements might be.

However, in the case that the experimental errors do belong to a normal distribution, the least-squares estimator is also a maximum likelihood estimator.[11]

These properties underpin the use of the method of least squares for all types of data fitting, even when the assumptions are not strictly valid.

Limitations

edit

An assumption underlying the treatment given above is that the independent variable, x, is free of error. In practice, the errors on the measurements of the independent variable are usually much smaller than the errors on the dependent variable and can therefore be ignored. When this is not the case, total least squares or more generally errors-in-variables models, or rigorous least squares, should be used. This can be done by adjusting the weighting scheme to take into account errors on both the dependent and independent variables and then following the standard procedure.[12][13]

In some cases the (weighted) normal equations matrix XTX is ill-conditioned. When fitting polynomials the normal equations matrix is a Vandermonde matrix. Vandermonde matrices become increasingly ill-conditioned as the order of the matrix increases.[citation needed] In these cases, the least squares estimate amplifies the measurement noise and may be grossly inaccurate.[citation needed] Various regularization techniques can be applied in such cases, the most common of which is called ridge regression. If further information about the parameters is known, for example, a range of possible values of  , then various techniques can be used to increase the stability of the solution. For example, see constrained least squares.

Another drawback of the least squares estimator is the fact that the norm of the residuals,   is minimized, whereas in some cases one is truly interested in obtaining small error in the parameter  , e.g., a small value of  .[citation needed] However, since the true parameter   is necessarily unknown, this quantity cannot be directly minimized. If a prior probability on   is known, then a Bayes estimator can be used to minimize the mean squared error,  . The least squares method is often applied when no prior is known. When several parameters are being estimated jointly, better estimators can be constructed, an effect known as Stein's phenomenon. For example, if the measurement error is Gaussian, several estimators are known which dominate, or outperform, the least squares technique; the best known of these is the James–Stein estimator. This is an example of more general shrinkage estimators that have been applied to regression problems.

Applications

edit
 
Least square approximation with linear, quadratic and cubic polynomials.

Uses in data fitting

edit

The primary application of linear least squares is in data fitting. Given a set of m data points   consisting of experimentally measured values taken at m values   of an independent variable (  may be scalar or vector quantities), and given a model function   with   it is desired to find the parameters   such that the model function "best" fits the data. In linear least squares, linearity is meant to be with respect to parameters   so  

Here, the functions   may be nonlinear with respect to the variable x.

Ideally, the model function fits the data exactly, so   for all   This is usually not possible in practice, as there are more data points than there are parameters to be determined. The approach chosen then is to find the minimal possible value of the sum of squares of the residuals   so to minimize the function  

After substituting for   and then for  , this minimization problem becomes the quadratic minimization problem above with   and the best fit can be found by solving the normal equations.

Example

edit
 
A plot of the data points (in red), the least squares line of best fit (in blue), and the residuals (in green)

A hypothetical researcher conducts an experiment and obtains four   data points:       and   (shown in red in the diagram on the right). Because of exploratory data analysis or prior knowledge of the subject matter, the researcher suspects that the  -values depend on the  -values systematically. The  -values are assumed to be exact, but the  -values contain some uncertainty or "noise", because of the phenomenon being studied, imperfections in the measurements, etc.

Fitting a line

edit

One of the simplest possible relationships between   and   is a line  . The intercept   and the slope   are initially unknown. The researcher would like to find values of   and   that cause the line to pass through the four data points. In other words, the researcher would like to solve the system of linear equations   With four equations in two unknowns, this system is overdetermined. There is no exact solution. To consider approximate solutions, one introduces residuals  ,  ,  ,   into the equations:   The  th residual   is the misfit between the  th observation   and the  th prediction  :   Among all approximate solutions, the researcher would like to find the one that is "best" in some sense.

In least squares, one focuses on the sum   of the squared residuals:   The best solution is defined to be the one that minimizes   with respect to   and  . The minimum can be calculated by setting the partial derivatives of   to zero:     These normal equations constitute a system of two linear equations in two unknowns. The solution is   and  , and the best-fit line is therefore  . The residuals are       and   (see the diagram on the right). The minimum value of the sum of squared residuals is  

This calculation can be expressed in matrix notation as follows. The original system of equations is  , where   Intuitively,   More rigorously, if   is invertible, then the matrix   represents orthogonal projection onto the column space of  . Therefore, among all vectors of the form  , the one closest to   is  . Setting   it is evident that   is a solution.

Fitting a parabola

edit
 
The result of fitting a quadratic function   (in blue) through a set of data points   (in red). In linear least squares the function need not be linear in the argument   but only in the parameters   that are determined to give the best fit.

Suppose that the hypothetical researcher wishes to fit a parabola of the form  . Importantly, this model is still linear in the unknown parameters (now just  ), so linear least squares still applies. The system of equations incorporating residuals is  

The sum of squared residuals is   There is just one partial derivative to set to 0:   The solution is  , and the fit model is  .

In matrix notation, the equations without residuals are again  , where now   By the same logic as above, the solution is  

The figure shows an extension to fitting the three parameter parabola using a design matrix   with three columns (one for  ,  , and  ), and one row for each of the red data points.

Fitting other curves and surfaces

edit

More generally, one can have   regressors  , and a linear model  

See also

edit

References

edit
  1. ^ Weisstein, Eric W. "Normal Equation". MathWorld. Wolfram. Retrieved December 18, 2023.
  2. ^ Lai, T.L.; Robbins, H.; Wei, C.Z. (1978). "Strong consistency of least squares estimates in multiple regression". PNAS. 75 (7): 3034–3036. Bibcode:1978PNAS...75.3034L. doi:10.1073/pnas.75.7.3034. JSTOR 68164. PMC 392707. PMID 16592540.
  3. ^ del Pino, Guido (1989). "The Unifying Role of Iterative Generalized Least Squares in Statistical Algorithms". Statistical Science. 4 (4): 394–403. doi:10.1214/ss/1177012408. JSTOR 2245853.
  4. ^ Carroll, Raymond J. (1982). "Adapting for Heteroscedasticity in Linear Models". The Annals of Statistics. 10 (4): 1224–1233. doi:10.1214/aos/1176345987. JSTOR 2240725.
  5. ^ Cohen, Michael; Dalal, Siddhartha R.; Tukey, John W. (1993). "Robust, Smoothly Heterogeneous Variance Regression". Journal of the Royal Statistical Society, Series C. 42 (2): 339–353. JSTOR 2986237.
  6. ^ Nievergelt, Yves (1994). "Total Least Squares: State-of-the-Art Regression in Numerical Analysis". SIAM Review. 36 (2): 258–264. doi:10.1137/1036055. JSTOR 2132463.
  7. ^ Britzger, Daniel (2022). "The Linear Template Fit". Eur. Phys. J. C. 82 (8): 731. arXiv:2112.01548. Bibcode:2022EPJC...82..731B. doi:10.1140/epjc/s10052-022-10581-w. S2CID 244896511.
  8. ^ Tofallis, C (2009). "Least Squares Percentage Regression". Journal of Modern Applied Statistical Methods. 7: 526–534. doi:10.2139/ssrn.1406472. hdl:2299/965. SSRN 1406472.
  9. ^ Hamilton, W. C. (1964). Statistics in Physical Science. New York: Ronald Press.
  10. ^ Spiegel, Murray R. (1975). Schaum's outline of theory and problems of probability and statistics. New York: McGraw-Hill. ISBN 978-0-585-26739-5.
  11. ^ Margenau, Henry; Murphy, George Moseley (1956). The Mathematics of Physics and Chemistry. Princeton: Van Nostrand.
  12. ^ a b Gans, Peter (1992). Data fitting in the Chemical Sciences. New York: Wiley. ISBN 978-0-471-93412-7.
  13. ^ Deming, W. E. (1943). Statistical adjustment of Data. New York: Wiley.
  14. ^ Acton, F. S. (1959). Analysis of Straight-Line Data. New York: Wiley.
  15. ^ Guest, P. G. (1961). Numerical Methods of Curve Fitting. Cambridge: Cambridge University Press.[page needed]

Further reading

edit
  • Bevington, Philip R.; Robinson, Keith D. (2003). Data Reduction and Error Analysis for the Physical Sciences. McGraw-Hill. ISBN 978-0-07-247227-1.
edit