Bifolia created using previous as reference

[npcompl33t]

Is it possible to do multivariate statistical analysis on multiple, different data sets at the same time? Ie. words analyis (your algorithm and/or LSA) plus margin distance plus text slant plus quimqu's results with bifolio hair/smooth sides plus...? (Just wondering... I guess there's a problem with assigning weights to each variable)

Its definitely possible to consider all of them together, with some weighting. The sequences i provided above actually did this with the golf MDS curves and the new data. Ultimatley i think to correctly sequence individual pages we will need several forms of overlapping evidence, because the uncertainty bands for any single piece in isolation are high. 

I don't think these markers provide a clear chain across the entire manuscript. There are several likely sequences, but also there are pages that don't seem to be connected. This could indicate a different production process was used for those sheets. 

For the chains that do appear though, it is now possible to compare different orderings against these new metrics. This is what i did for the Layfield and Davis comparison. This makes their 2nd proposal look much better than the golf ordering or their 1st proposal, which i think strengthens the case of that sequence over the other 2.