The experience of applying the effects of interaction:The study of the determinants of violence in straight couples
On the 6th of March, the first member of our Research and Study Group (RSG) Maria Rodionova made her own report at an open seminar. Maria introduced the other participants of the group and external listeners to results of creating a regression model with interaction effects on her own database. Application of the developed algorithm made the R squared more than three times higher, however, without including new variables.
Maria’s report was dedicated to the principles of choosing variables with the aim of building a model with high prediction power. In Maria’s own database this helped to increase a coefficient of determination (R squared) from 0,11 to almost 0,4, although, the topic of Maria’s research is very sensitive and devoted to determinants of violence in heterosexual couples. Thus, it was found that if the respondent has average income which is not controlled by his or her partner, this significantly reduces the presence of violence in a pair. At the same time, the opposite patterns are characteristic of those respondents, whose income is above average level and controlled by the partner at a dating phase. In this case, the level of violence in the pair increases. Seeing the sensitivity of the topic, final R squared can be considered quite high.
During the presentation, Alexey Rotmistrov, the head of the RSG, complemented Maria`s report with comments and additions. These helped the workshop participants understand more fundamentally a new regression modeling approach. Maria`s performance opened a series of seminars dedicated to the results of various techniques for selecting and accounting interaction effects in regression models. Soon, the other participants of the RSG will also show their results of the applied method developed within their own research directions.