Discriminant Analysis Models are very similar to regression models, but differ in one important respect. In Discriminant Analysis the key variable of interest Y is now categorised (eg, Buyer = 1, Non-buyer = 0) rather than being continuous like sales.
In this paper, a classification method (the index method) is proposed which aims to discriminate between members of two groups (e.g. buyers and non-buyers). The index method is compared with two-group discriminant analysis which is a standard and common technique of exploratory data analysis for selecting relevant variables and specifying the relationships between variables. It is shown how the index methods accounts for the problems met in applying two-group discriminant analysis.
This study was, therefore, designed primarily to determine the demographic and socioeconomic profile of do-it-yourselfers. Data obtained from 119 households located in the city of Niagara Falls and suburbs, and analyzed using discriminant analysis technique identified some important demographic and social variables. The discriminant function which was statistically significant was able to classify the majority of do-it-yourselfers into the appropriate group or category. The demographic and social profile of consumers determined in this study may be used by the do-it-yourself industry for developing and implementing successful marketing strategies.
For our initial evaluation we undertook a discriminant analysis to see if the age groups varied from one another with respect to attitudes towards living. The main objective of this phase was to compare the attitudes of the respondents towards living ten years ago to their attitudes today to see if identical age groups have changed their position in discriminant space, and when yes, then in which direction.
For our initial evaluation we undertook a discriminant analysis to see if the age groups varied from one another with respect to attitudes towards living. The main objective of this phase was to compare the attitudes of the respondents towards living ten years ago to their attitudes today to see if identical age groups have changed their position in discriminant space, and when yes, then in which direction.
Classical methods of analysis are ill-suited to dealing with this type of data, so that a new family of methods has recently been developed under impetus from C. Hayashu, J. de Leeuw, G. Saporta, M. Tenenhaus and F.W. Young. They attempt, by optimal quantification subject to inherent constraints, to determine the "best" factor analysis, the "best" regression procedure, and so on. This paper is concerned with a decisional (l.e. guiding selection of explanatory variables) method put forward by G. Saporta.
Commercial activity is no longer considered as having to apply to the entire population, but rather addresses itself to a clearly-defined sector which is given the name of "target consumers". It will at once be appreciated that a knowledge of the universal informative activity of these consumers will be an invaluable asset in order to adapt successfully the commercial activity to which the population is to be submitted. The means of communication, in particular, should take into consideration the following factor at various levels: the nature of the message to be communicated and the publicity backing.
Commercial activity is no longer considered as having to apply to the entire population, but rather addresses itself to a clearly-defined sector which is given the name of "target consumers". It will at once be appreciated that a knowledge of the universal informative activity of these consumers will be an invaluable asset in order to adapt successfully the commercial activity to which the population is to be submitted. The means of communication, in particular, should take into consideration the following factor at various levels: the nature of the message to be communicated and the publicity backing.
The conventional methods of analysing market research data are useful, but have their limitations. They consist, usually, of the construction and examination of one-way tables by characteristics like social class, age of housewife etc., together with the associated two-way and three-way tables. This approach leads, particularly with the aid of an electronic computer, to an overwhelming amount of tabulated results and it is doubtful whether these are inspected in any systematic way. The task which faces the research agency is to remove from the client the burden of having to examine a mass of tabulated results, whilst at the same time providing him with a guarantee that useful data has not been overlooked. Methods of isolating the most important aspects are discussed in this paper and applied to material collected on the ownership of consumer durables; the techniques are however applicable to a wide range of data. An application of the Belson method of matching population samples is considered, but this is also found to be limited to some extent for the present purpose. This is followed by a discussion and application of Discriminatory Analysis, in order to isolate the important characteristics underlying observed gradients in ownership. Simple but effective models are then fitted to ownership levels calculated for the various segments of the population indicated by the important characteristics.
The purpose for which discriminant analysis was developed, and for which it tends to be used in practice, is to obtain a function discriminating between individuals belonging to two or more populations, which can then be used to allocate further individuals to one or other of these populations when it is not known to which population these new individuals do, in fact, belong. The purpose of the discriminant analysis reported by Mr. Squirrell clearly differs from this, and further explanation of its purpose is asked for. In particular there seems to be the danger that an attempt might be made to "interpretâ the coefficients or weights of the discriminant function, despite the difficulties of doing so already referred to by Dr, Meer and pre-echoed in the paper itself.