Feedforward artificial neural network to estimate iq of mental retarded people from different psychometric instruments

The estimation of a person´s intelligence quotient (IQ) by means of psychometric tests is indispensable in order to determine possible mental retardation or intellectual disability based on the most common classification systems. With some subjects, however, it is not possible to use more complex tools such as the Wechsler scales, which are universally recognized as being the most reliable, in that they require minimum capabilities which are not always possessed by people affected by serious cognitive defects. This means it is necessary to use other psychodiagnostic tools that are better suited to a subject´s specific condition, but also that it is then necessary to reach a common metric so as to compare the reliability of the results obtained and thus ensure a homogeneous diagnosis. The concrete problem arising in the diagnosis of mental retardation using the IQ is thus the need to match the scores, obtained using different tests, with the Wechsler IQ, which is the most commonly used and universally recognized test for the diagnosis of degrees of retardation. In this paper we present the use of feedforward artificial neural networks (ANNs) to search for the best estimate of the Wechsler IQ provided by four different psychodiagnostic tools. To this end, a database was created, administering four different tests, besides the Wechsler scale, to the same group of mentally retarded subjects, in order to generate IQ estimation models on the basis of the scores obtained in the other tests via the use of ANNs. The results were then compared with other statistic modeling methods, in terms of accuracy and reliability.