Abstract
Matrix algebra is one of the essential tools of energy systems analysis. A grasp of even the most fundamental concepts allows much that would otherwise be extremely complex to be reduced to a few simple matrix equations; and the definition of most statistical tools, such as least squares regression, becomes extremely simple when the tedium of scalar algebra is replaced by the elegance of matrix expressions.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Preview
Unable to display preview. Download preview PDF.
References
This Chapter is designed solely as a brief refresher of the basic tools of matrix algebra, statistical analysis, and linear programming. We make no claim of comprehensiveness; indeed, the idea here is to present the minimum amount of information necessary for the comprehension of the techniques presented in subsequent chapters.
Transposition is sometimes denoted by use of a prime — thus A+ is the transpose of A.
The general definition of the determinant of an n-th order matrix A is given by as |A| = Σ ± a1α c2β… anu the sum being taken over all permutations of the second subscripts, with even permutations having positive signs, add permutations a negative sign. A permutation is said to be odd when the number of inversions is odd: an inversion, in turn, is said to occur when of, say, two integer subscripts the larger precedes the former. For further discussion, see e.g. Johnston (1963).
This, too, is an enourmous subject, to which many textbooks have been dedicated. Our purpose here is simply to review the basics, introduce those techniques most commonly encountered in the statistical analysis of energy economic phenomena, and to lay the groundwork for some of the more complex estimation problems (of energy demand elasticities, and the like), presented in Chapter 4. Any student familiar with this material can proceed to the next section since only fundamentals are discussed here (as they might be treated in any graduate level course on econometrics).
Minimizing the sum of the absolute values of the deviations leads to a linear programming problem (known as MAD estimation, for minimum absolute deviations). For an excellent introduction to this approach, see e.g. Rogers (1968). The MAD approach is sometimes useful in situations in which conventional least squares estimation is beset by serious problems, some of which we shall encounter later in these pages. The MAD technique has perhaps found greatest application in problems of mathematical demography.
We follow here the notational convention used in Johnston (1963)
The rules of matrix and vector differentiation are analogous to that of scalars: compare (2.36) and (2.37) with (2.25) and (2.27), and recall footnote 7, supra.
See e.g. Johnston (1963) p. 112.
This is the basic form of regression model in M. Chenery and M. Syrguin
(1975), although we have omitted here the non linear terms and time dummy variables for illustrative purposes.
The classic illustration of the difference between a type I and type II error is in terms of the axiom of anglo-saxon law: the presumption being that an accused is innocent (the null hypothesis) unless proven guilty (the alternative). The requirement for unanimity in juries is an effort to minimize type I error; under no circumstances do we wish to reject the null hypothesis if it is in fact true (i.e. we wish to minimize convictions of persons actually innocent) even if it means committing a large type II error (accepting innocence even though the individual may be guilty).
Also known sometimes as the region of significance.
This test is most commonly used to test for sample means where the sample size is small.
The value of t is tabulated in most statistics textbooks.
For a more leisurely exposition, but also based on a single numerical abstraction (of “Malawi”) see e.g. Todaro (1971), p. 87ff.
This “country” will be used for illustrative purposes throughout this book.
A convex space is one for which it is impossible to find a line drawn between two points on the boundary of the space that intersects another part of the boundary. For example, the space is not convex, since the line AB intersects the boundary
Subject of course to the exception of having an objective function that has the identical slope to one of the constraints: then the objective function will coincide with that constraint, and any one of the combinations of X1 and X2 that lie along the feasible segment of such a constraint will be optimal. For example, if in our example, the objective function were Max S = 4x1 + 8x2, then the optimum would be anywhere along the line YY of Figure 2.6.
For a full discussion of the numerical problems of computing inverse matrices, see Forsythe and Moier (1967).
But see exercise 4 for a discussion of significant figures.
Author information
Authors and Affiliations
Rights and permissions
Copyright information
© 1984 Springer-Verlag Berlin Heidelberg
About this chapter
Cite this chapter
Meier, P. (1984). Mathematical Fundamentals. In: Energy Systems Analysis for Developing Countries. Lecture Notes in Economics and Mathematical Systems, vol 222. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-642-48337-0_2
Download citation
DOI: https://doi.org/10.1007/978-3-642-48337-0_2
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-12879-3
Online ISBN: 978-3-642-48337-0
eBook Packages: Springer Book Archive