The purpose of this study aims to investigate the effects of IMPROVE (Introducing the new concepts, Meta-cognitive questioning, Practicing, Reviewing, Obtaining mastery, Verification, Enrichment and remedial) instruction on university students’ statistics achievement and attitude. IMPROVE is the state-of-the-art and metacognition-oriented instruction which is to enhance reasoning by training students to use a series of self-addressed metacognitive questions. Some literatures have confirmed the effects of IMPROVE method on achievement and affect. However, little is known about its effects on university statistics instruction. This study adopts qusi-experimental design, which consists of control group and experimental group, to compare the traditional instruction method and IMPROVE instruction. Results show that these two groups perform the same causal reasoning when they are on the baseline. After both of these two groups participate one semester of statistics course, results show the experimental group has higher statistics achievement than that of control group. Similarly, experimental group also has higher score in certain dimensions of statistics attitude than that of control group. These findings support that the IMPROVE method has positive effects on university students’ statistics achievement and attitude. Based on the conclusions, some recommendations and suggestions for future research are discussed.