Learning computational modeling after many years mastering statistical techniques required a paradigmatic shift in my thinking about time.
Time is fundamental to the study of development. In statistical models – growth curves, regression – I had thought of time as a variable that had (1) a direct effect on my outcome variable (legitimacy of parental authority) or might moderate the effect of other variables on my outcome variable (the association of parental warmth with parental authority might vary with age). But time was a variable that started at one point (e.g., age 12) and proceeded forward.
I quickly found that although the conceptual models I would draw for my statistical models and my dynamic systems models looked fundamentally the same – lots of constructs connect with arrows – they were not. In statistical models, time was either implied in a causal pathway (parenting style influences legitimacy beliefs over time, not simultaneously) or explicitly (age). In dynamic systems or computational models, on the other hand, time was iterative. I would start my process at time 1. My processes would occur at that step. Now I was at time 2 but beginning my process again at a new point.
During this time, I attended a series of workshops on computational modeling:
- Computational Modeling Oberlin College (6/2010, 6/2011, 6/2012, 6/2014, 6/2015)
- Dynamic Systems Modeling in Biological Sciences (Shodor 6/2009)
I was also very fortunate to be able to travel to the University of Groningen to work with Saskia K
unnen. I think I have memorized her excellent book introducing how to use computational modeling to better understand adolescent development.
Other readings were critical to me as well.