Stochastic methods of data analysis of marketing and social surveys

In the marketing surveys and applied sociology quantitative estimates are typically based on asymptotic characteristics of sample frequencies. To overcome the problem of heterogeneity of population «quota sampling» method that reflects the main categories of the population structure is commonly used. This paper presents a method of statistical data analysis of categorized sets based on random sampling. The method is based on conditional probabilities for statistics of binary relations on finite sets «observations — dichotomous factors». Compared to the quota based methods, the proposed approach significantly improves the accuracy of estimates for the population and allows obtaining estimates of frequencies for the categories of population for any given classification