31.2. Removing Nonstationary Means and Variances

If μ_t = E{z_t) depends on t, we can try to find a parametric model for μ_t, estimate the parameters of the model, and take to remove this source of nonstationarity. A popular approach is to parameterize fit as a linear function of some set of observed variables X_t = {X_1t,..., X_mt)′ and parameters β = (β₁,...,β_m)′, leading to