INTRODUCTION: Market surveys, consumers behavior, the questionnaire design and the survey methods. The statistical analysis of the purchasing behaviors. Market segmentation. Selection of the target market.
First part: DATA COLLECTION FROM SURVEYS The sampling design: population, sample, population parameters and their estimators. Sampling and non-sampling errors. Sampling space and inclusion probabilities. Sampling designs: simple random sampling, stratified sampling, cluster sampling, two-stage sampling, systematic sampling, probability proportional-to-size sampling. Estimation of totals and means. Direct, ratio, and regression estimators. Estimation for domains. Fixing the sample size. More complex sampling designs. Multipurpose surveys. Sample allocation methods: proportional allocation, univariate optimum allocation (Neyman), multivariate optimum allocation (Bethel).
Second part: ANALYSIS OF THE MARKET DATA Statistical analysis of the market data: basic notions of statistical inference, survey sampling, estimation, testing hyphoteses. Basic linear algebra. Data matrices. The analysis of the demand with linear models: linear regression, multivariate regression. Non-linear models. Explorative multivariate analysis: principal components analysis, biplots, correspondence analysis. Market segmentation: cluster analysis, binary and multiple segmentation, classification and regression trees. Analysis of variance (Anova, Ancova, Manova, Mancova). Linear mixed effects models: fixed effects models, variance components models, random effects models, multilevel models. Statistical spatial analysis with linear models. Diagnostics with linear models.
Third part: USING LINEAR MODELS IN SURVEY ESTIMATION Estimation using auxiliary variables. Model-assisted estimation: GREG estimators. Small domain estimation: direct and indirect estimators, composite estimators, synthetic and regression-synthetic estimators. Indirect estimation. Model-based estimation with linear mixed effects models. Estimation by linear predictors. Area-level estimation with the Fay-Herriot model. Mean squared prediction error of the best linear unbiased predictors.
