Welcome | Research | Teaching | Resources | Interests | Contact Me
Developmental processes cannot be understood outside of the contexts in which they occur. Positive and negative developmental trajectories occur from a constellation of influences and experiences that occur in cultural and historical contexts, leading substantive researchers to question whether generalizations about development should be made across individuals at all. In his “Manifesto on psychology as an idiographic science,” Molenaar (2004) proves that such generalizations depend upon the assumption that a developmental process is ergodic, an assumption that is almost always violated in studies of interindividual differences. First, in order to be ergodic a process must be invariant, or stationary, over time, meaning that neither mean levels nor variances can change over time. Second, in order to be ergodic, the process must be homogenous across participants, meaning that all participants must follow the same developmental model. If these assumptions are violated, then it is impossible to generalize from population findings to individual processes, meaning that population-level statistics such as t-tests, ANOVAs, and even multi-level models, have no relevance for any one individual.
The problems caused by violations of these assumptions are apparent in sexual orientation research. For example, many researchers have argued that linear coming out models, which were based on population-level reports, are generally inaccurate when describing any one individual’s coming out process. The coming out process is non-ergodic; it varies with circumstances and experiences that are unique to an individual’s life. In order to understand development, research must focus on the individual, a conclusion that seems obvious in many ways, but is controversial in practice as studies are currently ranked in importance by number of participants they include. The consequences of non-ergodicity cannot be ignored, however.
To address these issues, I am studying how theories of developmental processes relate to longitudinal data collection and analyses. For processes that are ergodic, I would like to examine how analyses of longitudinal data at the group-level, including latent growth curve modeling, survival analysis, and structural equation modeling, can be applied to the individual-level of development. For non-ergodic processes, I would like to examine the application of individual-level techniques such as time series analysis, hidden Markov modeling, and p-technique, including identifying individuals with similar processes for the purpose of aggregation.
As part of my work, I developed a website for the Developmental Systems Group. Here, you will find a longer description of ergodicity, listings of DSG members (who are all interested in the application of person-specific models), and downloadable programs with detailed instructions and examples. These programs include code for conducting hidden Markov models in R and an Extended Kalman Filter with Iteration and Smoothing created by Dr. Molenaar for estimating time-varying parameters for non-stationary time series. Please feel free to contact Dr. Molenaar (or me) with questions regarding DSG. All are welcome to join.
For links to some useful sites regarding statistics, please visit my Resources Regarding statistics page.
