This study was to analyze the effects of parental educational level (P.EDU), science attitude (ATT), and valuing science (VAL) on science performance after establishing gender invariance in a representative sample of the Taiwanese eighth grade population drawn from the 2007 Trends in International Mathematics and Science Study (TIMSS). The official TIMSS five plausible values in scoring science performance were used as outcome variables, while independent variables were collected from student questionnaires. The effects of P.EDU, ATT, and VAL on science performance were estimated by a multi-group structural equation model. Results from multi-group analysis supported gender invariance. After establishing gender invariance, a considerable effect was detected between P.EDU and Taiwanese eighth grade science performance that was indirectly related through ATT and VAL with the direct effects being stronger than the indirect effects, and the indirect effects of ATT being stronger than VAL in both genders. Prospective impediments to multi-group invariance are suggested to be utilized when comparing different groups.