In order to explore the impact of math anxiety and problem type on computational estimation and to investigate the problem-size effect in computational estimation, 388 participants was tested with Revised Mathematics Anxiety Rating Scale（R—MARS） and self-designed multiplication estimation questions. The results showed that 1) math anxiety has a significant impact on computational estimation in two different contexts; 2) the accuracy of estimation was significantly different in different contexts. In particular, the accuracy in pure digital context is remarkably higher than in word problem context; and 3) problem-size effect in computational estimation was found in word problem context, but not in pure digital context