dht2 does not understand uniform key function models

[erex]

Highlighted by workshop participant (working with dung surveys over short distances). dht2 fails with uninterpretable error message when handed a "fitted" uniform function.

MRE:

library(Distance)
data(ducknest)
units <- convert_units(ducknest_units$Units[2],
                       ducknest_units$Units[1],
                       ducknest_units$Units[3])
uni <- ds(ducknest, key="unif",
          convert_units = units,
          truncation = 2)
hn <-  ds(ducknest, key="hn",
          convert_units = units,
          truncation = 2)
gof_ds(uni)
gof_ds(hn)
trouble <- dht2(uni, flatfile=ducknest,
                convert_units = units,
                strat_formula = ~Region.Label)
notrouble <- dht2(hn, flatfile=ducknest,
                convert_units = units,
                strat_formula = ~Region.Label)

dht2 failure for uniform model

Error in abs(x) : non-numeric argument to mathematical function

generated by this line of code

df_unc <- lapply(ddf, varNhat, data = res)

but quite happy with the half normal model.

[erex]

supplement:

the above lapply function is calling varNhat. Suspect the problem is in varNhat which on this line is looking for the Hessian of the fitted model to get some variance bits-and-pieces.

There won't be a Hessian when the detection function is uniform without adjustments.

[LHMarshall]

as Eric pointed out above in the varNhat function line 50 fails as there is no hessian matrix

>   vcov <- solvecov(model$hessian)$inv
Error in array(x, c(length(x), 1L), if (!is.null(names(x))) list(names(x),  : 
  'data' must be of a vector type, was 'NULL'

> model$hessian
NULL

[lenthomas]

Note - the uniform is fitted with no adjustments. Forcing one adjustment makes the error go away.