Latent growth curve models are very useful for those of us conducting longitudinal trials, longitudinal randomized trials. And latent growth curve models allow us to examine the trajectory of growth on a specific variable over time. And in my case, since I' m looking at comparing different types of intervention programs, I' m looking to see if drug use patterns over time are different across groups. So growth curve models allow me to see are there different growth patterns over time and whether they favor one condition more than another intervention condition.

So there are other possible alternatives and other methods to use; a commonly used method when conducting longitudinal analyses is to do repeated measures analysis of variance. And although it's commonly used there are a lot of limitations with repeated measures.

For example, it does not handle missing data well. Some of these longitudinal trials you often have lost to follow up so repeated measures does not handle that missing data well. Another type of analyses, for example depending on the outcome in question is if you're interested in time to developing a certain disease, time to developing a certain condition, for example, an alternative method may be survival analyses which answers the question time to a certain event occurring. But for the type of studies that we use we work with, latent growth curve models are probably the statistic of choice.

Latent means it's unobserved, so often times we have an outcome that is not really an observed variable. So it's not something that we can go out and measure physically. So for example, if we talk about family functioning, family functioning is not something that our surveys capture. Our surveys tap in to different constructs or different indicators of family function. For example, communication, so how well do parents and children talk to each other. One of our other surveys may capture family support for the adolescent.

So although we tap into parent-adolescent communication, family support, parental involvement, all of those indicators are really tapping in to this latent or unobserved construct called family function.