An Evaluation of Estimative Uncertainty in Large Language Models