The building of experimental designs from observational data

This paper deals with optimal experimental design criteria and neural networks in order to build experimental designs from observational data. It aims at addressing three main issues: i) the introduction of two different approaches by the theoretical point of view; more precisely, T-optimal designs together with Generalized Linear Models (GLMs) and the approach through artificial neural networks; ii) the proposal of two algorithms, suitably defined in order to exploit existent data; iii) the comparison of the two suggested methods by means of a simulated case study in the technological field.