Showing papers in "The Statistician in 1984"

Measuring Skewness and Kurtosis

Abstract: The question of how to measure the degree of skewness of a continuous random variable is addressed. In van Zwet (1964) a method for ordering two distributions with regard to skewness is given. Here, using the concept of comparative skewness, we consider properties that a measure of skewness should satisfy. Several extensions of the Bowley measure of skewness taking values on (-1, 1) are discussed. How well these measures reflect one's intuitive idea of skewness is examined. These measures of skewness are extended to measures of kurtosis for symmetric distributions.

Statistical expert systems: design

Abstract: Introduction de la notion de systemes experts pour donner des avis statistiques. On examine la place de tels programmes dans le continuum de logiciel statistique, on decrit des processus cognitifs humains compris dans l'acte de donner des avis statistiques et on note un parallele saisissant entre diagnose medicale et consultant statistique

Exploring Multivariate Data using the Minor Principal Components

Abstract: The minor principal components of a random vector x are the orthonormal linear combinations of the variables in x which have minimum variance. This fact may be used to detect interrelations amongst the variables as well as to identify those variables which can be predicted with high accuracy from the others. Application of these ideas to the variable selection problem in multiple regression and to the identification of outliers in multivariate data are discussed.

Some Comments on Bivariate Regression

Abstract: We discuss the relationship between the multiple correlation coefficient and the correlations among the variables in a bivariate regression. We point out a common error concerning the variation explained by regression variables. A geometric interpretation is given and examples to illustrate the error are provided. example, the proportion of variation in y explained by x1 andx2 together can be greater than the sum of the proportions explained by x1 alone and x2 alone. It would appear that this situation may be misunderstood; in fact, two recent textbooks on regression actually state that this cannot arise. We present results, diagrams and examples which show that it does arise and that it may be quite common. We will call such an occurrence enhancement. The fitted regression coefficients can also exhibit bizarre behaviour. If we compare the fitted coefficient of, say, x1 in the regression of y on x1 with the fitted coefficient of x1 in the bivariate regression then these coefficients may differ by orders of magnitude or be of opposite sign. Such behaviour is sometimes called suppression. We show that there is a close connection between enhancement and suppression. We give examples taken both from bivariate and multiple regression to illustrate these phenomena.

The development and use of the NCB strategic model

Abstract: Description des elements du modele strategique du National Coal Board et de leurs correlations. Applications diverses de ce modele: prevision de la demande d'energie repartie par source de 1980 a 2000 en Grande-Bretagne, prevision de la demande industrielle de charbon en 2000 et 2020, prevision du commerce international du charbon en 2000 et prevision des couts marginaux du charbon en Europe Occidentale en 2000

Gas Demand Forecasting

Abstract: Description des differentes sortes de prevision de la demande de gaz effectuees par la British Gas Corporation suivant la periode de temps consideree

An alternative to the sign test in a matched pairs design

Abstract: This note explores the problem of ties in the sign test when the comparisons within pairs are based on ordered classes. It is suggested that under certain assumptions the number of ordered classes can be used to specify a sampling distribution which includes the ties: the resulting test may be more powerful than the orthodox sign test. If the assumptions are not met the test is, it is argued, conservative. This note explores a problem which arises when a matched pairs design of survey or experiment is combined with a level of measurement which only provides ordered classes. The matched pairs research design is widely used in medical and social sciences in an attempt to control for a multitude of factors which may disturb the outcome of an experiment or survey. The standard statistical texts recommend a sign test with the null hypothesis: P(Xa>Xb) =P(Xa Xb or Xa

Contribution of renewable energy technologies to future energy requirements

Abstract: Prevision de la production d'energie a partir des sources d'energie renouvelables en Grande-Bretagne pendant la periode 1985-2025 pour differents scenarios de developpement economique et d'augmentation des prix de l'energie produite par des techniques conventionnelles

Techniques for forecasting electricity demand

Abstract: Description de quelques-unes des techniques utilisees pour la prevision de la demande d'energie electrique en Grande-Bretagne et du procede par lequel les previsions sont etablies

Load research and its application to electricity demand forecasting

Abstract: This paper describes the actual systems for half-hourly demand metering of consumers, together with the scope of the research programme in covering different types of consumer and electricity usage, and the statistical methodology involved. Demand patterns obtained from the research may be linked with sales (TWh), actual or estimated, to build total system load curves. This approach, which automatically compensates for differing mixes of consumers or sales, is demonstrated by preparing a mediumterm (seven year) forecast of system maximum demand. As an aid to evaluation the "forecast" method is applied to historic data and is seen to provide an excellent explanation of changes in the relationship between sales and demand (expressed in terms of unrestricted load factor) which have occurred over the last 15 years. The method may be applied from any base year and is readily able to absorb new or improved data whilst also providing an indication of where future research could most usefully be directed.

Power Aspects of Split-plot Designs

Abstract: The sensitivity of power in analysis of variance to the departures from the in-built assumptions is discussed in Kanji (1975) and Kanji and Liu (1983) and in obtaining the power they have used the general linear model. The purpose of this paper is to discuss a particular case of the above situation, namely the split-plot design. The results obtained indicate that although the power value is seriously affected by the heterogeneity of whole-plot error variances, the actual type I error is little affected. Also the power value is little affected when the errors are serially correlated within subplots in the whole-plot treatment comparisons. It is also shown that heterogeneity of subplot error variances affects the power value without influencing substantially the type I error.

Forecasts and Public Policy

Abstract: Formulation de sept questions clefs a l'occasion d'une evaluation des methodes de prevision energetique a long terme aux Etats-Unis et en Grande-Bretagne. Influence de l'action gouvernementale. Presentation d'approches pour la validation des modeles et etude critique des modeles econometriques. Recommandations pour la construction de modeles

A note of caution on using statistical software

Abstract: En reordonnant les entrees de donnee d'un programme de regression fait d'ensemble de programme, on a observe des differences significatives dans les resultats. Ceci est utilise pour illustrer l'extension de dommage fait par les erreurs numeriques generees dans les calculs

Estimating the Ultimate Economically Recoverable Hydrocarbon Reserves of the North Sea Province: Phase. 1 The Theory

Abstract: Presentation de la methode utilisee pour l'evaluation de reserves ultimes d'hydrocarbures economiquement recuperables de la Province de la Mer du Nord

Energy Statistics and the Planning Perspective

Abstract: Revue des differentes difficultes d'etablissement de previsions en matiere d'energie a partir de donnees statistiques