This article analyzes the test-retest reliability of network size and density measures originally implemented in face-to-face surveys, but implemented in this research in a telephone survey for examining “Social Relationships among Czech Citizens.” For this purpose we can divide family and friendship networks. Network size is measured with the number of family members; respondent’s friends at work, in the neighbourhood, as well as other friends. Network density is operationalised as frequency of contact with family members and friends. The analyses show a high test-retest reliability of the network size and density measures of the family network, but very low reliability for the measures of the friendship network. More detailed analysis reveals that the low reliability of network size, density of friends at work, and contact frequency with friends, can be explained in terms of respondent characteristics. In contrast, the low level of reliability exhibited by other variables is independent of respondent characteristics., Julia Häuberer., and Obsahuje bibliografii a bibliografické odkazy
Advances in the statistical analysis of longitudinal data has been so rapid, that it has been difficult for empirically oriented social scientists to remain informed of all new developments in this important area of social methodology. This article offers some guidance on the use of various types of panel data analysis techniques, paying particular attention to the analysis of longitudinal panel data. The aim of this article is to describe in a succinct manner the logic underpinning a number of panel analysis techniques; outlining the types of inferences that can be drawn from employing specific techniques, and providing the reader with references to the literature associated with particular forms of panel data analysis. Five types of panel data analysis are discussed: Event history analysis, Sequential analysis, Hierarchical linear (or multi-level) modeling (with application to longitudinal data analysis), Structural equation modeling with longitudinal data, and use of Log- linear and Markov chain models for longitudinal data with categorical variables., Petr Pakosta, Petr Fučík., and Obsahuje bibliografii a bibliografické odkazy