The quandary of diagnosing mathematical difficulties in a generally low performing population

Abstract

ABSTRACT. Brazilian students’ mathematical achievement was repeatedly observed to fall below average levels of mathematical attainment in international studies such as PISA. Objective: In this article, we argue that this general low level of mathematical attainment may interfere with the diagnosis of developmental dyscalculia when a psychometric criterion is used establishing an arbitrary cut-off (e.g., performance<percentile 10) may result in misleading diagnoses. Methods: Therefore, the present study evaluated the performance of 706 Brazilian school children from 3rd to 5th grades on basic arithmetic operations addition, subtraction, and multiplication. Results: In line with PISA results, children presented difficulties in all arithmetic operations investigated. Even after five years of formal schooling, less than half of 5th graders performed perfectly on simple addition, subtraction, or multiplication problems. Conclusions: As such, these data substantiate the argument that the sole use of a psychometric criterion might not be sensible to diagnose dyscalculia in the context of a generally low performing population, such as Brazilian children of our sample. When the majority of children perform poorly on the task at hand, it is hard to distinguish atypical from typical numerical development. As such, other diagnostic approaches, such as Response to Intervention, might be more suitable in such a context.

Random Forest multiclasse: a diagnostic study of mathematical learning difficulties

“…We observe an opportunity for studies whose approach is carried out with multiclass machine learning methods, where the comorbidity influence and diagnose heterogeneity (HAASE et al, 2012; HABERSTROH; SCHULTE-KORNE, 2019) can be characterized. Also, psychometric cut-offs are recognized to be a support diagnosis instrument that requires joint clinical evaluation, (MAMMARELLA et al, 2021;WEISS, 2020;GOMIDES et al, 2021) what were not clearly approached on available studies. Moreover, when applying a diagnosis specialized protocol, one expects the dimensionality to increase, and data accommodates numerical and categorical results.…”