The marginal distances

[andreastausend]

There are several distinct comparisons possible. We exemplify the idea with the example for "pain" from the ticket explaining the matching procedure. Then the marginal distance could look like this

mdist_(pain)(pain1, pain2)= (w_pain* dist_typeofpain(type1,type2)+ w_location* dist_location(location1,location2)+ w_tissue * dist_tissue(tissue1,tissue2)+...)/(#particles)

Most of these distances could be the trivial metrik. For the location a 2-dimensional map of the body could be used, that fits into a ball of radius one. The distance in that case would be the euclidian distance of the coordinates. For the type of pain medically more similar types of pains could have a smaller distance than 1. The triggers of pain (always there, only when moved, only when strained, always but strain matters, less when moved) could also be allocated in a graph with different length of axises. For instance it seems reasonable to implement the function such that

dist_trigger(only when moved, only when strained) < dist_trigger(only when moved, less when moved)

The metriks for "trigger" and "typ of pain" could be represented by a matrix.

What needs to be done:

1. specify for each type of finding a meanigfull comparisson, which leads to a numerical quantification of similarity between 0 and 1.