Time-series data of population density or biomass occasionally contain missing values. In such incomplete-data cases, we cannot directly estimate population parameters (e.g., growth rate) using a ML (maximum likelihood) estimator constructed for complete-data sets. For incomplete-data from discrete-time Markov chain models, I studied an iterative method to calculate ML estimates based on an EM (expectation–maximization) algorithm and an MCMC (Markov chain Monte Carlo) method. The EM procedure at the kth step goes as follow: (1) Calculate the conditional expectation of the log-likelihood of complete-data given the incomplete-data observed and the current (kth) value of the parameters. Because it is often difficult to calculate analytically the conditional expectation of the log-likelihood, use Monte Carlo integration to compute it. For the Monte Carlo integration, generate random numbers following the probability density function of the missing values using an MCMC method. (2) Compute k+1th parameters maximizing the conditional expectation of log-likelihood of complete-data. As an application of the method, I estimated the birth rate of bluegill sunfish using the incomplete catch-per-unit-effort (CPUE) data from Potoka Lake, Indiana. In conclusion, this method provides a way of estimating parameters of a wide class of Markov chain models for natural populations from incomplete time-series data.