The recent adaptation of a Chinese parent version of the Autism Spectrum Rating Scale showed the Modified Chinese Autism Spectrum Rating Scale to be reliable and valid for use in China. The aim of this study was to test the Modified Chinese Autism Spectrum Rating Scale for fit to the Rasch model. We analysed data from a previous study of the Modified Chinese Autism Spectrum Rating Scale which comprised 1593 non-cases and 420 children diagnosed with autism spectrum disorder. We used super items based on groups of locally dependent items and item deletion when necessary to achieve good fit to the model for each of the three subscales identified by Zhou et al. and for the full 59-item Modified Chinese Autism Spectrum Rating Scale. The resulting conversion tables enable the use of genuine unidimensional, interval level scores for the total score and three subscales. Reliability was high with Person Separation Index values ranging from 0.83 to 0.89 for the three subscales and 0.79 for the total scale. In addition, we were able to identify a full-scale version of the Autism Spectrum Rating Scale and its three subscales that are all free of differential item functioning in relation to the five person factors recorded namely age, sex, caseness, relative and city. In future studies, the Teacher version of the Modified Chinese Autism Spectrum Rating Scale needs examination with Rasch analysis.