Skip to content. | Skip to navigation

Navigation

RSS feed

Statistics seminars

Videos from statistics seminars

14.2.2020 Mediation analysis for associations of categorical variables: The role of education in social class mobility in Britain - Jouni Kuha (London School of Economics)

48:51
We analyse levels and trends of intergenerational social class mobility among three post-war birth cohorts in Britain, and examine how much of the observed mobility or immobility among them could be accounted for by existing differences in educational attainment between people from different class backgrounds. We propose for this purpose a method which quantifies associations between categorical variables when we compare groups which differ only in the distribution of a mediating variable such as education. This is analogous to estimation of indirect effects in causal mediation analysis, but is here developed to define and estimate population associations of variables. We propose estimators for these associations, which depend only on fitted values from models for the mediator and outcome variables, and give variance estimators for them. The results of the analysis show that the part that differences in education play in intergenerational class mobility is by no means so dominant as has been supposed, and that while it varies with gender and with particular mobility transitions, it shows no tendency to increase over time.
3.3.2020

28.2.2020 Fast tensorial independent component analysis - Joni Virta (University of Turku)

39:21
A novel method of tensorial independent component analysis is proposed based on TJADE and k-JADE, two recently proposed generalizations of the classical JADE algorithm. The new method achieves consistency and the limiting distribution of TJADE under mild assumptions, and at the same time offers notable improvement in computational speed. A trade-off between computational speed and assumptions is controlled by a tuning parameter which has a natural interpretation as the maximal kurtosis multiplicity. Simulations and timing comparisons demonstrate the method's gain in speed and show that the desired efficiency is obtained approximately also for finite samples. The method is applied successfully to large-scale video data, for which neither TJADE nor k-JADE is feasible. Finally, an experimental procedure is proposed to select the values of a set of tuning parameters. Joint work with Niko Lietzén, Pauliina Ilmonen, and Klaus Nordhausen.
4.6.2020