Evaluating the effect of a pedagogical innovation is often done by looking for a significant difference in a content measure using a pre-post design. While this approach provides valuable information regarding the presence or absence of an effect, it is limited in providing details about the nature of the effect. A measure of the magnitude of the pre--post change, commonly called learning gain, could provide this additional information to chemical education researchers. In this paper, we compare two methods of measuring learning gains using data from large-scale administrations of the Chemical Concepts Inventory at four universities. The intent of this study is to compare various measures of learning gain, not to contrast the teaching effectiveness at the four universities. In this gain analysis, we introduce a method based on Rasch modeling and discuss the advantages offered by this type of analysis over more commonly used measures of learning gain.