Hello again,

No I didn't envisage a binary threshold as I think they would be highly limitating. Logically it should be be a function of the range of correlation coefficients each dataset returns. So for example, with a data that returns lots of correlations at above say 0.8 then the threhold should be quite high (e.g 0.4), with a data set that returns very few high correlations or none at all, the threshold should be lower so that I have enougth observations to analyse for that dataset.

I was thinking that the easiest way would be to have a line in the code where I can manually change the threshold based on what the correlation matrix is showing for each data or if this is automatically built into the code then have an option of a manual overide so I can see all the correlations if I want to. But you probably have a better idea of the best way to do it. The key thing for me is something with a degree of flexibility so that I can change the criteria easily if I need to.

Thanks,

Nick