search

gadget-framework / gadget3

TMB-based gadget implemtation
GNU General Public License v2.0
8 stars 6 forks source link

Order dimensions correctly for surveyindices #143

Closed lentinj closed 4 months ago

lentinj commented 4 months ago

@vbartolino would you mind giving this branch a go at some point and let me know if it helps fix your example #142?

If it helps, then I'll merge this.

Fixes #142 (well, hopefullly)

vbartolino commented 4 months ago

It does the job, great! Note that the two data extractions below both seem to work now. min and max lengths in the model are 3 and 21 cm but in the first case the data component length span 0:Inf. Are the two extractions really equivalent for the model? Should one of the two be preferred?

> si.ven.aco <- mfdb_sample_count(mdb, c('age'),
       list(...,
            age = mfdb_group('age3pl'=3:maxage))) [[1]]
> dimnames(g3_distribution_preview(si.ven.aco))
$length
[1] "0:Inf"
$age
[1] "2:10"
...
> si.ven.aco <- mfdb_sample_count(mdb, c('age','length'),
       list(...,
            age = mfdb_group('age3pl'=3:maxage))) [[1]]
> attributes(si.ven.aco)$length$all <- c(minlen,maxlen)
> dimnames(g3_distribution_preview(si.ven.aco))
$length
[1] "3:21"
$age
[1] "2:10"
...
lentinj commented 4 months ago

Are the two extractions really equivalent for the model?

No, the former ignores length, whereas the latter filters the plus group.

A plus group is added by default to a stock:

> s <- g3_stock('s', seq(3, 21, 3))
> s__num <- g3_stock_instance(s, 100)
> s__num
length
   3:6    6:9   9:12  12:15  15:18  18:21 21:Inf
   100    100    100    100    100    100    100

Aggregating this into a "3:21" vector removes this plus group:

> s_noplus <- g3_stock('s_noplus', c(3, 21), open_ended = FALSE) ; s_noplus
g3_stock s_noplus with dimensions:
$length
[1] "3:21"

> g3_eval(~stock_iterate(s, stock_reshape(s_noplus, stock_ss(s__num))))
[1] 600

OTOH, a length group of "0:Inf" includes everything:

> s_inf <- g3_stock('s_inf', c(0,Inf), open_ended = FALSE) ; s_inf
g3_stock s_inf with dimensions:
$length
[1] "0:Inf"

> g3_eval(~stock_iterate(s, stock_reshape(s_inf, stock_ss(s__num))))
[1] 700

Should one of the two be preferred?

Ignoring the plus-group isn't ideal, but depending on your model the results of doing so may be insignificant.

Instead of removing length, you could also set the open_ended option in mfdb_group. The data returned would be the same, but the option will follow through to gadget3 and result in "3:Inf".