need to check if the Distance helpers work when there are levels in the distance model (i.e. for veg type)

[techisdead]

I recall there being an issue at one stage with retrieving w and edr from more complex distance models. Just need a second eye / brain on these functions and maybe a test or two for when the Distance model has variables (like veg type)

[techisdead]

w won't work but this is an example that does (edr example only but can be done for both). @pyandell99 or @symbalex please check my reasoning and I'll go and code it up

library(Distance)
library(collision)


#https://distancesampling.org/Distance/articles/covariates-distill.html
#


data(amakihi)

ds_model <- ds(amakihi, transect="point", key="hr", formula=~OBs,
                  truncation="15%")

# ds_model$ddf |> str()
# ds_model$ddf$fitted |> names()
# 
# ds_model$ddf$meta.data$int.range  #w is ok
# ds_model$ddf$data$OBs |> head()
# 
# ds_model$ddf$data |> class()
# 
# covs <- amakih$OBs

covariate_col <- "OBs"
## Check and only work if covs are factors
if(
  any(
    sapply(ds_model$ddf$data[, covariate_col], function(x) {
      inherits(x, "factor")}) != TRUE)
) stop("I can only work with covariates that are factors")

df_res <- data.frame(fitted = ds_model$ddf$fitted,
           ds_model$ddf$data[, covariate_col] ) |> unique()
           
df_res$edr <- df_res$fitted * w_from_distmodel(ds_model)

df_res[, c(covariate_col, "edr")]



## Example with two covariates
## 

amakihi |> head()
amakihi |> str()

## make second covariate for this test
amakihi$HAS <- factor(amakihi$HAS)

ds_model <-  ds(amakihi, transect="point", key="hr", formula=~OBs+HAS,
                truncation="15%")


covariate_col <- c("OBs", "HAS")

ds_model$ddf$data[, covariate_col] |> unique()

df_res <- data.frame(fitted = ds_model$ddf$fitted,
                     ds_model$ddf$data[, covariate_col] ) |> unique()



df_res$edr <- df_res$fitted * w_from_distmodel(ds_model)

df_res[, c(covariate_col, "edr")]

[symbalex]

@techisdead line 32 is funky. Otherwise it's fine

[symbalex]

nah I have a better way

library(data.table)
dt_res <- as.data.table(ds_model$ddf$data[, covariate_col] |> unique())
# point transect
dt_res[
  , c("fitted", "se.fit") := as.data.table(predict(object = ds_model
                                                 , newdata = .SD
                                                 , esw = TRUE
                                                 , se.fit = TRUE))
]
# point transect
dt_res[, edr := sqrt(fitted/pi)]
# line transect
# ....