The story begins with training experiments on clumsy children showing that certain procedures have particular effects (Sims et al, 1996a and b). Two measures are used: TOMI and KS, which behave slightly differently in response to different training procedures.
These data were modelled together with some develpmental data. Altogether there were 4 types of child: normal, clumsy, very clumsy and young; 4 types of training and the 2 outcome measures.
Three internal variables were the minimum, one specific to each of the outcome tests and one which contributed to both. On that basis we attempted to model the data set using 2 or 3 levels for the internal variables. while this could be done by hand, COGENT seemed an obvious way of checking it. The basic model could be set up in a couple of hours, entering the specification for subject group and training programme by keyboard. Fiddling around with automating a sweep through all groups and all training regimes took longer and Richard. This enabled a faster and complete turn round in adjusting the parameters of the system. Ability to interface the program with a display to illustrate what is happening would be wonderful.