Poster #69 - Comparing the Relationship Between Early Reading and Mathematics Skills Across Performance Levels Using Quantile Regression

• In Event: 3-106 - Poster Session 11
  In Poster Session: PS 11 Section - Education, Schooling

Integrative Statement

Previous research has indicated that mathematics and reading skills are highly related at early grades. However, little research exists whether this relationship differs by performance level across time. Examining differences in factors affecting performance between low and high achieving students can help provide insight to policy and teaching approaches which are targeted to specific ability levels. This study explores possible differences in the relationship of mathematics and reading achievement between students of varying mathematics ability beyond simply the overall trend for all students.

The study uses data from the Early Childhood Longitudinal Study Kindergarten Class of 2010-2011, a nationally representative sample of kindergarten students tracked across five time points from fall of kindergarten to spring of third grade.
At each time point, vertically scaled direct cognitive assessments were individually administered for reading and mathematics. Student demographic characteristics such as gender, race/ethnicity, and family socio-economic status (a derived variable based on parental education, parental occupation, and household income) will be included in the final analysis. A variable that measures the teacher’s rating of the student’s approach to learning is also included to measure the relationship between child’s approach to learning and mathematics performance.
Quantile regression analysis is used to estimate regression coefficients of specific quantiles in mathematics achievement to model the relationship between predictors (e.g. reading scores, demographics) and high and low math performing students and any differences between the two groups. This is modeled across five-time points from grades K to 3 to investigate any difference in the relationship between mathematics and reading from kindergarten through third grade.

Figure 1 plots the quantile regression coefficients at grade K fall and grade 3 spring. The y-axis represents the size of the coefficient of the labeled covariate, while the x-axis represents the quantile (e.g. 0.5 represents the median). Comparison between corresponding graphs across time points illustrate the change in the relationship between quantile and coefficient. Table 1 presents the regression estimates for OLS and quantile regression at the .10 and .90 levels. Preliminary analyses indicate that the effect of reading performance and approach to learning on mathematics performance are overall positive across all grades and ability levels; however, in fall of kindergarten, the difference between the coefficient of reading scores at the bottom 10th percentile versus the top 90th percentile was significantly negative (.558 vs .829, F=64.31, p<.001), whereas this difference was significantly positive by spring of third grade (.863 vs .523, F=75.64, p<.001). A similar effect was found for student’s approach to learning. Meanwhile, the opposite effect was found for demographic variables such as Black and Hispanic, going from a positive difference in kindergarten to a negative difference in third grade.
These findings provide further insight into the connection between early childhood mathematics performance and reading performance. Looking at multiple quantiles in mathematics scores reveals differing gradations in the size of the main effects. Moreover, when looking across time points, the direction of the difference in some effect sizes between high and low math performers may change over time.

Authors

• Thomas Sun, American Institutes for Research
  Presenting Author

• B. Jasmine Park, American Institutes for Research
  Non-Presenting Author

• Xiaying Zheng, American Institutes for Research
  Non-Presenting Author

• Yuan Zhang, American Institutes for Research
  Non-Presenting Author

• Mengyi Li, American Institutes for Research
  Non-Presenting Author