One of the most important uses of statistical analysis is to investigate the associations or relationships between different variables. Understanding these relationships is of importance to an investigator for several reasons: It helps in the understanding of the phenomenon or phenomena under investigation. It gives insights into possible causal mechanisms between the variables. It is an important step in the construction statistical models which relate the variables to each other. These models may be used to improve the quality of predictions. The study of association in Statistics falls into two broad areas: 1. Correlation between two or more variables, and 2. Association between two or more categories in a frequency table.
The simplest form of analysing data is to form survey tabulations. This is done by counting the number (and percentage) of people that fall in to the predefined categories of our questionnaire. The basic tool for the survey analyst is the cross tabulation in which one or more questions on the questionnaire form the rows of the cross tabulation and one or more different items form the columns. The simplest form would be where one question forms the rows and one demographic forms the columns.
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.
It is rarely possible to study the effect of several marketing factors simultaneously. New media, however, may provide such opportunities. Teleshopping does so, because here sales is linked directly to product, price, advertising, media plans, etc. During four weeks of July, 1990 the Danish Television Channel 2 used most of its advertising time on a special campaign in order to assess the possibilities for teleshopping in Denmark. Twenty-three widely different products were marketed in this period, and the number of telephone calls for each product were registered. In total, about 16 calls were registered during the four-week period. On this background the television channel assesses the campaign to have been a success. This data material, consisting of daily registrations of the number of calls for each product, 460 observations in total, is the basis for our analysis of the number of calls. We propose a statistical model that describes 90 percent of the variation in data, making the number of calls a function of 1) a product-specific level, 2) the number of spots, 3) the number of viewers, 4) the time of day, 5) the day of the week, 6) the time of the month, and 7) price.
From 1989 the National Readership Survey in Denmark is conducted by using the FRY-method and the CATI-technique. The reading probabilities are a very essential part in the FRY-model. The paper describes how they are calculated and how they are corrected if they are illogical. The questions used for calculation of the reading probabilities are the frequency-question and the FRY-question. As the number of FRY-readers are of decisive importance for the statistical margins of the coverage it is necessary to calculate the reading probabilities on the results from the largest possible sample size. It is presented how the statistical margins of the coverage are reduced essential by calculating the reading probabilities on a yearly basis and on media group levels. Finally the paper illustrates how the calculation of duplication, sole readership and non-readership are more complicated by using reading probabilities. But the consequences of using probabilities instead of traditional 0-1 variables are much more realistic results.
Goelette II, which succeeds to Goelette I, is a sales forecasting system for dispatching at the time of each issue MARIE-CLAIRE Group magazines into the distribution network of the "Presse en France". The primary objective of the system is to keep an economically acceptable percentage of unsold copies without risking being out of stock organisation (N.M.P.P.). It is also an instrument of analysis permitting, through the use of seasonal coefficients, the grading in a given situation of a serie of published issues, and by the calculation of deviations between forecasts and actual achievements to put forward hypothesis concerning the influence of factors external to the model (coverage, competition, date of publication, etc.). Finally, it is a management tool permitting the execution of statistical analysis and controls necessary for the conduct of business.
The main thesis of this paper is that we have taken a new step toward understanding democratic elections since the early seventies. We are now in a position to understand the influence of processes of public opinion on voting behavior. The new influences are the climate of opinion; the use of a "quasi-statistical sense" people have to assess which attitudes are on the increase among the general public and which are on the decrease; the willingness to testify to one's voting intention in public or the tendency to keep silent ("spiral of silence"); the threat of isolating supporters of the other side by imbuing election themes with a moral dimension, and the role of the media, which are by definition public, in this process; and, again, as already treated by Lazarsfeld/Berelson/Gaudet but not by subsequent election research: the role of opinion leaders, the two-step flow of communication and the bandwagon effect.
This paper discusses current and future developments in the ways pre-classification can be used. The paper particularly examines the use of small area statistics linked to other data bases. A suggestion is also made as to how pre- classification can be used, to optimally stratify and allocate samples, when the estimation of more than one variable is considered.
This paper deals with efficient estimation of descriptive statistics such as means, totals and percentages from rotating panel surveys (panels with partial replacement). It is supposed that the variable under study is measured repeatedly. When the correlation between consecutive panel estimates of that variable is high and positive, a considerable gain in precision can be achieved using special methods. This gain in precision can be translated into a smaller panel size if the accuracy of the estimates is satisfactory already. The gain can be achieved by using a so-called composite estimate , which is a weighted combination of current and previous estimates. The method of the composite estimate will be described in section 2 and its variance will be discussed briefly in section 3. In section 4 some results will be given on the possible gain in efficiency which the use of composite estimates may yield when respondents are replaced after three measurements. Most of the theory in sections 2 and 3 however holds for all panels with replacement after a fixed number of measurements, whether it be two or ten waves. The paper continues with a worked-out example in section 5 and the conclusions are given in section 6.
Finding samples representing segments of the general population when the rate of strike (percentage of the segment in total population) is inferior to 20%, is becoming a common performance asked from the practitioner by client marketing companies. This is at least the experience of SECED - Research International in France. The survey of practice within this agency indicates clearly that not only targeted samples of the population, but very narrow target samples form an increasing part of the commissioned work of this agency. Special techniques, not resorting to the well established statistical sampling techniques, are currently used to cope with this problem. They are presented in decreasing order of statistical orthodoxy, from the "telephone screening" to the "introduction technique", through use of data-files and statistical inference, with remarks on their cost efficiency. None of these techniques is actually original, and the paper is mainly intended to initiate discussion and sharing of experience on the practical ways and means of finding these "oiseaux rares".
SISDATA, the Statistical Information System, designed and maintained by Slamark International, an Italian information system and marketing consultancy firm, deals with secondary statistical data, that is, the data of the main sectors of a country's economic and social activity normally surveyed in official statistics. The system has two facets: a socio-economic analysis function applied to extremely disaggregated data covering a considerable time span, and a dynamic trend observer dealing with the most recent aggregated data in the form of time series. Currently envisaged in a European dimension, the Italian archive is already operative and will be joined by the French, Federal German and British sectors. The System is, moreover, multi-lingual, the descriptions to the statistical tables (rows and columns) being provided in the language of the country surveyed and English as the interface language.