Corrigendum: On-line Expectation–Maximization Algorithm for Latent Data Models

The unnumbered equation near the top of page 603 contains some typographical errors and the statement that immediately follows it is also incorrect. These should be replaced by

Solving equation (28) is easy for a well-specified model, i.e., when π = g_θ*, as the FIM I_obs(θ^*) that is associated with the (observed) data model then satisfies

When ζ = 0, the solution of the Lyapunov equation is thus given by $Σθ*=Iπ−1θ*/2$⁠. The model being well specified also implies that I(θ^*) = I(θ^*), and, hence, the asymptotic covariance matrix is given by half the inverse of the complete data FIM in this case.