This study examines variables of student math proficiency relevant to family, teacher, and school characteristics, and investigates how effects of significant variables vary by geographic areas in the State of Missouri, United States, using aggregated data at the school-district level. Ordinary Least Squares (OLS) regression analysis derives a statewide model of five significant predictor variables with an R square of 0.35, showing that school districts with higher family income, higher parent education levels, higher two-parent percentage, more teacher years of experience, and higher percentage of certified teaching assignments tend to have higher student math proficiency. Geographically Weighted Regression (GWR) local regression analysis using the five significant variables from OLS regression achieves R squares between 0.37 and 0.45 at different bandwidths, i.e., the spatial extent of local regression. Examining GWR local regression coefficients, we found that the variable effects can change considerably across different geographic areas, and some prominent variable effects can only be manifested at a small GWR bandwidth. The results of this study identified the most influential factors of math proficiency for local areas, which provide useful insights for policy makers to prioritize educational resources for improving student math proficiency.