1
$\begingroup$

For example, I am interested in specific brain regions' development over time.

I have data, but data is one-time point data. it is not longitudinal.

But in the data, I have participants aged from 0~17.

Also, I have the depression score of each participant.

So, my research question is:

if there is any statistically significantly different in certain brain regions' development trajectories between depressed children and healthy children.

But my data is collected at one-time point, and it is not longitudinal.

So I an concerned that if I can use longitudinal data analysis methods to compare developmental trajectories between two groups. (although sample size is enough, and participants are ranged from 0~17)

$\endgroup$

1 Answer 1

0
$\begingroup$

A general way to evaluate changes over time is to model your measure of brain-region development flexibly as a function of time, for example with a regression spline, and include an interaction term between that and your measure of depression. Significance of the (set of) interaction terms indicates whether there is a difference associated with depression. That basic strategy is used for longitudinal data in generalized least squares, as outlined in Chapter 7 of Frank Harrell's Regression Modeling Strategies.

That strategy, however, is not restricted to longitudinal data; it can be used for anything measured as a function of time. To some extent this is even simpler than a longitudinal generalized least squares model, as for any one brain region there are no intra-individual correlations to take into account.

The potential problem is that the variability in results among individuals might limit your ability to detect a true difference related to depression. That's similar to the possible difference of power between a 2-sample t-tests and paired t-tests.

You don't say how many brain regions you are measuring, whether you have specific hypotheses about one or more brain regions, or whether you are just looking for any brain regions that you might find to differ. With multiple regions in the same individual, you should take correlations among regions within the same individual into account. If you are evaluating many regions, you need to adjust for multiple comparisons.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.