Jul
27

July Updates: Actor Partner Interdependence Modeling

The past month has been full of exciting progress. I have been busy conducting follow-up sessions with the adolescents in our longitudinal sample, working on my literature search and introduction, and learning a new type of statistical analysis.

One of the core assumptions underlying most common statistical analyses is that of independence, which means that one participant’s scores do not affect the level of another participant’s scores. Because the participants in my sample came into lab for the first time in pairs with their best friends, I do not have independence in my data. Therefore, I cannot use any of the types of analysis that I have used before. Instead, I’m using a type of analysis called Actor Partner Interdependence Modeling (APIM), which accounts for similarity between dyads. Using this, I can look at how socialization processes function in adolescent friendships while accounting for the similarity in behavior and traits that drives adolescents to become friends in the first place.

This analysis is a bit complicated to understand, but luckily for me, a former student in my lab wrote out a detailed guide for how to conduct it. I also am reading the book that first formally outlined this type of analysis (Dyadic Data Analysis, 2006) and reaching out to former students for guidance. While learning a new type of analysis can be frustrating at times, it is exciting to have a new tool to explore my data.

Comments

  1. It’s really interesting that you’re diving so deep into how your longitudinal data was collected. I think a lot of social scientists tend to make overly generous assumptions about data collection to access the statistical procedures they’d like to use, so a push in the other direction is smart. To what extend does this correction impact the types of analysis you’re able to do?