Mathematics learning performance is a key stage in educational program planning. Mathematical performance appears as the result of complex variables and multi-factors. The literature on mathematics performance described several factor s that might lead to learning difficulties (Silver, Pennett, Black, Fair, & Balise, 1999). The first factors are related to the basic capabilities of children define their performance levels (e.g., IQ). This is difficult to treat and resistant to educational interventions. Secondary factors for mathematics learning performance are related to (1) individual variables that can be influenced/changed (match anxiety, math attitudes, self-efficacy; (2) the background variables that related to the personal family socioeconomic-status (SES) and other family environment variables. (3) The instructional environment (e.g. nature of the mathematics method, quality of educational interventions, time investment, use of tools). There are a lot of theoretical models to explain the factors on the performance (e.g., Bennet, 1978; Carroll, 1963; Glaser, 1976). For the first dimension, most of the studies focused on the relationship between student’s skills and belief (e.g., motivation, attributions, regulation, participatory behaviours and engagement, self-concept) and mathematics performance (DiPerna & Elliott, DiPerna, Mroch, & Lang, 2004). While age, gender, ethnicity and other factors differences were statistically significant, they were not as prominent as the differences noted for socioeconomic status (SES) (Reyes & Stanic, 1998; House, 1992; Fan, 2001; Rothman & McMillan, 2003). The meta-analysis on socioeconomic status (SES) and academic achievement between 1990 and 2000 revealed that a medium to strong SES–achievement relation, which was moderated by the unit, the source, the range of SES variable, and the type of SES–achievement measure. The relation is also contingent upon school level, minority status, and school location (White, 1982; Sirin, 2005). For the second dimension, the families’ SES was similar to the economic development level of the region and the school which may cause the student non randomly distributed. And the meta-analysis studies result by Bosker and Witziers (1996) showed that there was 18% of the variance in performance between schools. And as the multilevel studies showed that if the class level was taken into account, the influence of school shrinked to small level (Creemer & Scheerns, 1989). Such as the effective learning time, structure teaching and opportunity to learn (Scheerens & Bosker, 1997), were the factors on the class level (Scheerens, 1992). Thus, this present study aimed to (a) examine the students’ performance of mathematics across school and grades and (b) identify the individual factors and his/her background factors and the environment factors which are responsible for the variation among students’ performance. Specially, to address (a), we focused on four issues: (1) the performance of the mathematics in primary school. To address (b), we examined: (1) the effects of individual difference on the mathematics performance; (2) the characteristics of family difference which contributing to the mathematics performance; (3) the school effects over and above individual difference and family difference on the mathematics performance.

Multilevel analysis is the most appropriate method for modelling the interaction effects of variables at the different levels (Bryk & Raudenbush, 1992; Hox, 2002). In multilevel research, the data structure in the population is hierarchical, and the sample data are views as a multistage sample from this hierarchical population (Hox, 2002). The two-level hierarchical linear model was used to examine the independent association between the individual-level variables and the school-level variables and the mathematics performance. Dependent variables: Mathematics performance level was measured by the IRT model. Independent variables: (1) School-level variables including GDP, region and teacher variables; and (2)Individual-level variables including the demogolical variables such as(a) grade, (b) age, (c) gender, (d) race, (f) hand priority, (g) psychometric intelligence variables and Family variables (F): It includes the (a) Socio-economical status (SES) .

In this study, we explored the effect of the school and students in China on mathematics learning performance at the end of the school year. The result indicated that in null model 80.07% of the variance in mathematics performance is situated at the student level and 19.93% at the school level. There are other interesting discoveries in the research. The null model reveals that the estimated overall school average on mathematics performance is 0.4241 and is statistically significant. The estimated variance in the intercept is 0.2117 and the estimated value of the residual is 0.8507. All the predictors were centered before they were added into the model which means they have the means of zero. For the second model, we add grade as the fixed predictor. The estimate of the school mathematics performance is 0.4296 when the value of the predictor grade is zero. For each one unit increase in grade the expected value of mathematics performance increase by 0.2135 units.