We present iterative turbo-like equalizers for multiple-input multiple-output (MIMO) orthogonal frequency-division multiplexing (OFDM) systems subject to frequency-selective channels with rapid time variation. The proposed equalizers are obtained from a Bayesian approach based on the unbiased minimum mean-squared error (MMSE) criterion, but also admit a probabilistic data association (PDA) interpretation. Two equalizers are discussed: a basic version that makes use of symbol-level variance estimation, and an enhanced version that takes advantage of bit-level variance estimation. Simulation results show that both the presented equalizers outperform a previously proposed MMSE decision-feedback equalizer (DFE) of comparable complexity.