Visreg

[Nathan-Hebert]

What values does visreg use for the spatial/spatio-temporal random effects? Is there a way to specify the "time" variable to visreg?

Thanks!

[seananderson]

It would be turning them off / setting them to zero. I.e. setting re_form = NA in predict.sdmTMB.

sdmTMB/R/predict.R

Lines 954 to 968 in 1dc2801

	check_visreg <- function(sys_calls) {
	visreg_df <- FALSE
	re_form <- NULL
	se_fit <- FALSE
	if (any(grepl("setupV", substr(sys_calls, 1, 7)))) {
	visreg_df <- TRUE
	re_form <- NA
	if (any(sys_calls == "residuals(fit)")) visreg_df <- FALSE
	# turn on standard error if in a function call
	indx <- which(substr(sys_calls, 1, 10) == "visregPred")
	if (length(indx) > 0 && any(unlist(strsplit(sys_calls[indx], ",")) == " se.fit = TRUE"))
	se_fit <- TRUE
	}
	named_list(visreg_df, se_fit, re_form)
	}

(The internal code to make visreg work is kind of crazy.)

You could also consider now using ggeffects::ggpredict() instead.

sdmTMB/R/fit.R

Line 405 in 1dc2801

#' ggeffects::ggpredict(fit, "depth [all]") |> plot()

https://github.com/strengejacke/ggeffects/blob/master/tests/testthat/test-sdmTMB.R

There's a half-finished vignette that I should get out.

Or ggeffects::ggeffect(), assuming you aren't working with smoothers, which will be fast.