According to the teachings presented herein, a method and apparatus for communication signal processing advantageously use a mix of parametric and non-parametric correlation estimation in communication signal processing. Non-parametric estimation generates an “overall” correlation estimate for a received communication signal, and parametric estimation generates a “component” correlation estimate. The component correlation estimate is removed from the overall correlation estimate to form a partial correlation estimate that is used to process the received communication signal at least initially, such as in a pre-equalization stage. The overall and component correlation estimates are generated as impairment and/or data correlation estimates.