Non-linear growth curve model

[Shuping Lim]

Hi everyone,

I am new to growth curve modelling on STATA. The purpose of my study is to examine if prenatal maternal mood predicts the trajectory of child depressive symptoms from age 9 to 13, hence I want to use the depression variables measured across the 3 timepoints to model a growth curve for each sex.

These are the variables for my study:

1. Depression_age9 (continuous variable)
2. Depression_age10 (continuous variable)
3. Depression_age13 (continuous variable)
4. Sex
5. Prenatal maternal mood

Looking at the mean of the depression scores of each timepoints, the depression scores are lowest at the second time point, so I want to do a non-linear growth curve model for my analysis. How can I extract the rate of change in trajectory for each participant on STATA?

To do a non-linear growth curve model, I've converted my dataset to a long data format and keyed in the following code:

mixed depression sex age age2 || subjectid: age age2, covariance(unstructured) mle

age: age of the participant
age2: squared of age

I want to do a regression analysis to see if prenatal maternal mood is associated with the variability in the rate of change in depression across the 3 timepoints.

Greatly appreciate it if anyone can help me with this, thank you!

[Joseph Coveney]

It's still a bit confusing just what you're after.

First it's "to examine if prenatal maternal mood predicts the trajectory of child depressive symptoms".

And then it's "to see if prenatal maternal mood is associated with the variability in the rate of change in depression".

Assuming that your objective is more like your first statement, I wouldn't go about it by trying to "extract the rate of change in trajectory for each participant" (are you asking for the second derivative of the timecourse?) and using that as an outcome variable fitted to prenatal maternal mood as the explanatory variable.

Instead, you could fit a model with (centered) age as a predictor and include interaction terms involving prenatal maternal mood using factor variable notation. Something like

summarize age, meanonly
generate double cag = age - r(mean)
mixed depression i.sex i.pmm##c.cag##c.cag || subjectid:

if prenatal maternal mood is categorical, and

mixed depression i.sex c.pmm##c.cag##c.cag || subjectid:

otherwise.

I recommend keeping the model simple at least at first, omitting random slopes. The interaction terms in the fixed-effects equation will address the question of interest.