Dynicam linear regression, lags produce NA values

Question

How do I perform a dynlm regression when I want to included lags of both the dependent and independent variable.

For example with the following data;

a <- 1:10
b <- 5:15
a <- ts(a)
b <- ts(b)

dynlm(a  ~  L(a, 1:3) + L(b, 1:2))

What I am trying to obtain is a linear regression of the model where a depends on 3 of its own lags and 2 lags of b. However, I get NA values

I think I do not properly understand how dynlm works. Anyone who can give me some insight to what is going wrong?

Answer 1

With respect to the data shown in the question, the intercept and first lag give a perfect prediction so the other lags are not needed. Note that resid(dynlm(...)) is all zero.