Parse cut-derived factors in likelihood data.frames

[lentinj]

The main reason for the MFDB attributes is to specify the groupings that aren't represented in the data, i.e. you're trying to group lengths into seq(0, 100, by=10), but there's only data for 30..50. In this case we should be comparing e.g. [60, 70) to 0.

The various MFDB aggregates will put this into attributes, but the more R way of doing this would be cut:

> cut(c(10, 15, 25, 38), seq(10, 50, by = 10), right = FALSE)
[1] [10,20) [10,20) [20,30) [30,40)
Levels: [10,20) [20,30) [30,40) [40,50)

Parse factor strings as generated by cut() in likelihood_data, so we one can use dplyr group_by(cut(...)) instead of MFDB.

NB: dbplyr doesn't return groupings as factors, so this won't work there.

[lentinj]

@bthe we talked about this a few months a go, I think it will be useful for my Math4Fish script output as a simpler way of grouping the output data, so I've got on and done it.

[lentinj]

This should be done with the above, however a dplyr example would be much more intelligable:

ldist.lln.raw |> group_by(
  year = year, age = age,
  length = cut(length, breaks = seq(10, 100, by = 10), right = FALSE)
) |> summarise(number = sum(number))

To do this, we have to pull in dplyr as a suggests, a bit of a sledgehammer but probably worth it.

[lentinj]

Also:

What happens if you would do by_year = c(1981, 1999:2004, 2010) ? It would be cool if this would be wrapped into:

if(year >= 1981 && year < 1999) 
par.1981
else if(year==1999)
par.1999
....
else if(year >=2004 && year < 2010)
par.2004
else
par.2010

[lentinj]

As noted in https://github.com/gadget-framework/modelwizard/issues/6, taking factors of cuts is problematic, as they don't sort alphanumerically. We need to preserve the initial order when generating a factor