From appendix table S1, we can see that the attriting students (the majority of which are dropouts) differ from non-attriting students in baseline characteristics.Specifically, attritors were more likely to be male (column 2, significant at the 10% level), older (column 3, significant at the 5% level), have parents who are not at home (column 7, significant at the 1% level), have lower math scores(column 9, significant at the 1% level), and have lower computer scores (column 10, significant at the 10% level). Although there is imbalance between attriting and non-attriting students, the imbalance does not appear to bias our results. In particular, we estimate Lee Bounds (that account for problematic attrition) for the endline math and endline computer achievement outcomes (see our robust-ness check subsection 3.1.2 below).
求助翻译，尤其是加红字体部分。不胜感谢