Hi
I'm learning on splines. Trying to establish if my spline model is better than my 'orginal' model - which has showed non linear effects with LOESS
However I would like to plot a graph as shown in post 12 - can anyone help with the code? As I would not only like to use LR test but graphically visualise it.
Also, I would then like to test for interactions,between var2#var3.
What is your expert advice after determining if my spline model is better or not from the LR test - For example, let's say the cubic model is better, should I then test for interactions after testing the cubic model ?
I would use the code below.
Source: https://www.statalist.org/forums/for...ood-ratio-test
I'm learning on splines. Trying to establish if my spline model is better than my 'orginal' model - which has showed non linear effects with LOESS
Code:
//Cubic spline model (less restrictive model) stcox cspline* var1 var2 var3 estimates store m1 // Linear model ('more restrictive' model) stcox var1 var2 var 3 estimates store m2 // LR test to compare (m2 nested in m1) lrtest m1 m2
Also, I would then like to test for interactions,between var2#var3.
What is your expert advice after determining if my spline model is better or not from the LR test - For example, let's say the cubic model is better, should I then test for interactions after testing the cubic model ?
I would use the code below.
Code:
//Cubic spline model (less restrictive model) stcox cspline* var1 var2 var3 estimates store m1 // Interaction model stcox var1 var2 var3 var2#var3 estimates store m2 // LR test to compare (m2 nested in m1) lrtest m1 m2
Source: https://www.statalist.org/forums/for...ood-ratio-test
