Month names collection / metastudy

[eggyk]

I wonder if the many octembres that weren't added to the sheet (because their other months aren't interesting) follow the same trend?

How much this is sampling bias as Koen indicates is another question. We'd have to collect all instances of those variants regardless of their score of other month names. Maybe with enough samples we could find a regional trend?

My criteria for adding an entry has effectively been: "are the month names french?". Once I find something and see "janvier" (or anything non Januarius), I start filling in the manuscript's details before even looking at the rest of the month names. If I see latin names, I flick through to make sure they are all latin before disregarding. There are many times that I start filling in the names and it's just another boring set. In that case it goes something like: "mars, auril, may, juing, juillet, aoust, septembre, ooo octembre, nouembre, decembre". 

If a set has octembre or octobre, it will be in the data regardless of the other months. 

Edit to add: I would always compare octembres in absolute numbers, since the number of Octobres is basically infinite.

Honestly, the amount of Octobres actually isn't that high near the end of the 14th century. Octembre is - as of the current data - far more common than Octobre in that range. There is certainly an infinite amount of the latin "October", though. Here's a graph with the absolute numbers (which are still too low honestly, so more data will be helpful)
   

But I think with the broad dating ranges, such graphs can be misleading. How do you handle entries like 'early 14th century' or '15th century'?

"Early 14th century" is set as 1301-1325 normally, but I think for the purposes of the octembre graph I put "Early 14thC" to 1310 for ease. This is part of the reason that there are graphs for 30yr groupings and 20yr groupings, as most of these entries will find themselves in roughly the right place. 

There are actually 2 sheets specifically to deal with the imprecise "Xth Century" dates. One sheet omits them, and the other sets them to the middle of the century (1350, 1450). The stacked data really didn't change that much with them included, but the absolute numbers obviously see a peak at 1450. At some point I also used the "latest possible date" instead, and came up with similar results, I think.

Eventually I will plot this data more thoroughly; this was mainly to see if the trend I thought I saw was actually there.