I’m off to a workshop at The Methodology Center at Penn State University tomorrow. Two days of intensive learning about latent class analysis (LCA) and latent transition analysis (LTA). These are methods of finding hidden subgroups among a population of interest, for example teens at higher risk for diabetes, drug abuse or risky sexual behavior. It can be though of as factor analysis, but on people instead of variables. Each person has a probability of being a member of a group. Latent class analysis helps identify the groups at a single time point based on a high dimensional pattern of survey responses, and latent transition analysis allows the group membership to change across time. Adolescent behavioral researchers likely have identified certain characteristics of adolescents that put them in a high risk group, but how does the risk change over time? Is there a tipping point where students go down one path or another and don’t return?
The last research project I worked on had a combination of panel data (students surveyed once annually, a different cohort each year) and longitudinal data (students tracked and surveyed across years). This made for an interesting set of analyses, one that I tried using this LCA methodology on before, but ran into problems. It got pushed to the side for a more, “easier” method of categorizing students. Namely by a limited pattern of responses. If they said “A”, “B” and “C”, they are considered “high risk”. I was waist deep in my dissertation at that point, and being the only analyst/data manager on this small research team I didn’t have the spare time to figure out why the LCA/LTA models were not running.
I am really looking forward to this workshop! Part of the second day will be me getting to dig back into that data set, and have a real person who helped develop this methodology to help me debug, troubleshoot and understand the limitations of the model. Perhaps I was just trying to ask it something nonsensical and wasn’t aware of it.
On a totally different note, when I get back I want to try out the 7 minute workout. These have been going around for a while, but one of the artists for a web comic I read has recently returned, and he wrote a post about it that made it really seem effective, and more importantly, doable. Especially since it seems i’m busier now that i’m out of school than when I was in! Doesn’t help that there are three abstracts due in 2 weeks, when both weeks between now and July 9th are half weeks for me (workshop then holiday).
Will try to post thoughts from the workshop!
