Graphical exploratory analysis of vague data in the early diagnosis of dyslexia

Luciano Sánchez, Ana Palacios, M. R. Suárez, Inés Couso

We are in the initial stages of the design of a fuzzy rule-based testing system that can be used by unqualified personnel while screening the children for dyslexia. The main novelty of our work is the exploitation of low quality data (incomplete items, intervals, lists, subjective values, etc.) A sample of infants has been obtained, and some tests have been applied to them. In addition, a psychologist has examined and diagnosed each child. We want to relate the responses to the tests with the expert judgement of the professional, and highlight those factors that are involved in the early development of the dyslexia during the preschool age. However, this data being imprecise, we lack tools to assess the quality of each test and its influence in the diagnosis. We have used different graphical visualization techniques, to detect the most useful sets of factors, but have found that none of the approaches that we are aware of is able to show all the relevant information in the sample. Therefore, we propose a new Multidimensional Scaling algorithm, that gains a better insight into the spatial properties of the data and also into the amount of vagueness in the pieces of information comprising it.

PDF full paper