Right way to run one step sampling from stan_glm

I’m using stan_glm function in rstanarm package. But I just want to use stan_glm to sample a set of parameters \theta_1 in each iteration, but sample other parameters \theta_2 by my own sampler. Basically, what is the “correct” way to do it?

Is the following step right? In k-th iteration:

  1. sample \theta_1 by stan_glm:
    \theta_1^{(k)} = stan_glm(…,chains = 1, iter = 2,init = \theta_1^{(k-1)}); ## or should I use iter = 1?
  2. sample \theta_2 by my own sampler.