The non-decision time, in the response's own units

Without ndt_group() this is frm_linpred(fit, dpar = "ndt", type = "response"), which already reports a time. With ndt_group() the ndt link is a plain logit on a FRACTION of the row's own bound, so predict() reports that fraction and this multiplies it back out. Written so that a caller need not know which of the two a fit is.

Usage

ndt_time(object, newdata = NULL, ...)

Arguments

object

A fitted wiener(), lba(), rdm() or wiener_gng() model, or a fit from another package whose family carries a bound from ndt_bound_attach().

newdata

Optional data frame. It must carry the model's ndt_group() column when the model has one.

...

Passed to stats::predict(); re_formula and allow_new_levels are the useful ones.

Value

A numeric vector, one non-decision time per row.

Details

The bound is the fastest response of the row's ndt_group(), or of the whole data set when the model has no ndt_group(), or the max_ndt the family was given. Whichever it is, it is a property of the data the model was FITTED to, so a prediction on new rows is scaled by the same bound the fit used.

It reads the bound record rather than the family's name, so a fit from ANOTHER package whose family carries a bound through ndt_bound_attach() is reported here too, and frmtmb.learn::rlddm() is the first of those.

See also

wiener() for what the two parameterizations are and why there are two, ndt_bound() for the seam another package's family reaches them through.

Examples

set.seed(1)
d <- ddm_simulate(300, mu = 1.2, bs = 1.5, ndt = 0.25)
fit <- frm(bf(rt | dec(upper) ~ 1, bs ~ 1, ndt ~ 1, bias = 0.5),
           family = wiener(), data = d)
# with no ndt_group() the two agree: `ndt` is already a time
head(frmtmb::frm_linpred(fit, dpar = "ndt", type = "response"), 3)
#>         1         2         3 
#> 0.2569868 0.2569868 0.2569868 
head(ndt_time(fit), 3)
#> [1] 0.2569868 0.2569868 0.2569868