Dear Stata experts,
I'm handling with a longitudinal data from a twin study and want to investigate if parental history of myopia would affect myopia progression in offspring. In this dataset, I categorized parental refractive status into 5 groups and plotted the trend of myopia progression with increasing age using lowess smooth method. The plot is as following:
Array
It seems that the trend of progression showed a non-linear growth, so I tried to fit the quadratic growth model. In considering that the sample is from twin study, which means the individual is not independent from each other, so I chose to create a 3-level model:
mixed auto_r_se_ c.age15##p_c2 c.age15_2##p_c2 || fid: age15 age15_2 || id: age15 age15_2 , cov(unstruc)
in which variable age15 is centralized age around 15, age15_2 is age15*age15, p_c2 is the group variable.
The results are showed below:
And my questions are:
1. Is this code right for my purpose of investigating the difference of growth trend among groups?
2. How can I interpret the coefficient of the interactions, especially for those with age*age?
0 Response to Quadratic Growth Model
Post a Comment