"The Currier languages revisited" revisited

[MarcoP]

Ah, thank you. So Scribe 4 wrote both A and B pages, and there is no simple explanation for the two clusters. That's more interesting and more like the Voynich (nothing is ever simple).

[ReneZ]

Scribe 4 wrote all the pages with circular diagrams (except IIRC f57v).

[Bernd]

Potentially dumb question but how would PCA look like in 3 dimensions?
Do we get any additional resolution of clusters compared to the usual 2D projection?

[Koen G]

Potentially dumb question but how would PCA look like in 3 dimensions?
Do we get any additional resolution of clusters compared to the usual 2D projection?

I like this idea - certainly some (python?) method must exist to generate a cube that can be rotated on a webpage or something like that.

[bi3mw]

I like this idea - certainly some (python?) method must exist to generate a cube that can be rotated on a webpage or something like that.

You are not allowed to view links. Register or Login to view.

[RobGea]

Here is a cool tool that rotates graphs on a webpage,  but i just found it , i don't know anything about it, apparently just can just upload your data file in csv format and Presto !
PCA 3D Visualiser
You are not allowed to view links. Register or Login to view.

[Rafal]

These 3D plots look cool but are totally unreadable for me.

For 3 dimensions you could actually use three 2D graphs with AB, AC and BC dimenions on each.

[bi3mw]

I just did a test. The 3D plots generated with Python can actually be rotated in all directions.

[oshfdk]

These 3D plots look cool but are totally unreadable for me.

For 3 dimensions you could actually use three 2D graphs with AB, AC and BC dimenions on each.

For me even 2D plots are unreadable. I can see the clustering and patterns, but I'm completely clueless about the implications, most of the time.

I mean, suppose folio A and folio B are correlated under metric C (of whatever the proper lingo is). So what?

[Rafal]

I mean, suppose folio A and folio B are correlated under metric C (of whatever the proper lingo is). So what?

It means that they are similar on some dimension.

PCA (Principal Component Analysis) takes a lot of dimensions and reduces them to a smaller number of dimensions, let's say 2, although you can choose any number.

You lose some information by that and it is an important question how much you lose.

Imagine having two original dimensions for people - height and weight. You decide to reduce them to one dimension. What you get is one dimension telling if someone is "big" or "small". A man who is very tall will be "big" and a man who is very fat will be "big" too. But you lost some details.

In our Voynich analysis people took frequency of some letters bigrams. You have, let's say, 40 frequencies of most common bigrams for each page. They make 40-dimensional vector of original values. Such stuff is impossible to imagine visually but perfectly possible to work with it numerically. Then you reduce it to 2 dimensions.

These new dimensions are some abstraction and usually don't have any nice name in English language. You can just check them how they are "loaded" with your original dimensions which are frequency bigrams.