Possible to make computations on matched arrays in unequal matrices w/ missing values??

[_hs_]

Dear all,

I don't know if this is at all possible, but:

Does anybody know if there is a way to make computations involving two arrays (rows) in two different matrices/panels, where:

- the rows are not sorted i.e. values on an observation that should be matched/compared might be in different rows of the two panels
- the matrices are unequal i.e. values on an observation might only exist in one panel - or in none of them
- the row-ranges that are there contain missing values, notably in the first columns that should be ignored in the calculation in both ranges´
- the row-ranges are of unequal lenght - however, if this is a problem this is the easiest issue to solve - they could be made to match in lenght (number of columns).

I wish to compute summary statistics on the ranges, e.g. mean error, mean absolute approximation error, mean relative error etc.

Attached is an example which should give the (messy) picture:-) There is an attempted formula, but I know it sucks totally - but it might give an idea of the purpose.

Hope somebody will be able to help!

Thanks!

[_hs_]

Any ideas would be much appreciated - also if just to say that this is too difficult:-)??