Probability Pyramiding, Research Error and the Need for Independent Replication