Hello,

I have a OLS cost model with the cost of health care as the dependent variable and a series of 'need' and 'supply' variables as independent variables. The sample is vary large (over 50,000,000) observations and 800+ independent variables.

One of the need variables is age, and the model was developed using age splines with 18 knots:

mkspline asp0 1 asp1 5 asp2 10 asp3 15 asp4 20 asp5 25 asp6 30 asp7 35 asp8 40 asp9 45 asp10 50 asp11 55 asp12 60 asp13 65 asp14 70 asp15 75 asp16 80 asp17 85 asp18 = age_variable, marginal.


I have been asked to illustrate in a graph the consequence or benefits of using splines as opposed to age dummies (quinary age bands). While I can explain the difference and potential benefits, I am not sure how to illustrate it, especially with so many observations. Ideally I guess, you would graph the shape of the linear predictor against age for the model with splines and then graph the linear predictor for the model with dummies against age and see the difference. Not sure how to do this though or if it is correct.

has anybody had any experience of this before?