• A
  • A
  • A
  • ABC
  • ABC
  • ABC
  • А
  • А
  • А
  • А
  • А
Regular version of the site

News

Deep study of log-linear analysis and the first acquaintance with the R-package

The second meeting of our Research and Study Group (RSG) took place on the 6th of February and was dedicated to log-linear analysis and its application for searching interaction effects in social data. The meeting was held under the leadership of Svetlana Zhuchkova, the deputy head of the RSG, and lasted for an hour and a half. The very next day, on the 7th of February, at the initiative of the head of the group Alexey Rotmistrov there was held an additional closed seminar which gave an opportunity for participants to repeat previously studied material. Also, there were discussed the first results of the individual project, presented by Maria Rodionova.

Svetlana's report was devoted to the possibilities and limitations of log-linear analysis for searching interaction effects between categorical variables. In the presented material there were discussed various options for parameterization ("effect coding" and "dummy coding") and the possibility of their implementation in the statistical packages SPSS and R. The participants were also told about two ways of interpreting the results of the log-linear analysis. In the first case, it was proposed to proceed to the interpretation of significant micro effects within the framework of the unsaturated model, and in the second case just to stop the analysis on significant standardized residuals. At the end of the seminar participants shared their impressions of the applying the new method and checked results of the homework.

The next day, with the aim of consolidating the studied material, members of RSG were invited to continue calculating the interaction effects of the third and fourth levels manually by using the Excel. After that, Maria Rodionova presented the first results of her individual project, so that, other group members could be able to help her with the improvement of the quality of the regression model. Finally, after summing up the results of two last meetings, the head of the RSG Alexey Rotmistrov thanked all the participants and told about the following steps to continue the joint productive scientific work.