mayer79
hstats fails with cryptic error: String '_:_' found in grid values at unlucky position.
First of all, many thanks for this great package!
I'm currently trying to apply hstats to a causal forest (grf) object. However, as soon as I attempt to compute 2-way interactions (i.e., pairwise_m > 0), the function fails with an error that seems a bit cryptic.
From what I can tell, it is due to a hardcoded check—but it's unclear to me why this happens and how to work around it. I’d really appreciate it if you could point me in the right direction.
# wrapper for grf to work with partial_dep
pred_fun <- function(object, newdata, ...) {
predict(object, newdata, ...)$predictions
}
H <- hstats(cf_grf, X = X, pred_fun = pred_fun, verbose = TRUE, approx=TRUE)
Error in .compress_grid(grid = grid): String '_:_' found in grid values at unlucky position.
Traceback:
1. hstats(cf_full, X = X, pred_fun = pred_fun, verbose = TRUE, approx = TRUE,
. n_max = 100)
2. hstats.default(cf_full, X = X, pred_fun = pred_fun, verbose = TRUE,
. approx = TRUE, n_max = 100)
3. mway(object, v = v2, X = X, pred_fun = pred_fun, w = w, way = 2L,
. verb = verbose, ...)
4. wcenter(pd_raw(object, v = z, X = X, grid = X[, z], pred_fun = pred_fun,
. w = w, ...), w = w)
5. pd_raw(object, v = z, X = X, grid = X[, z], pred_fun = pred_fun,
. w = w, ...)
6. .compress_grid(grid = grid)
7. stop("String '_:_' found in grid values at unlucky position.")
Currently, I can reproduce only what is implemented:
If feature values contain the string "_:_", an internal function would fail. Thus, it yields an explicit error.
grid <- data.frame(X = c("", "", "_:_"), Y = c("_:_", "_:_", ""))
hstats:::.compress_grid(grid)
Could you please add a minimal reproducing example based on your setting?
I don't know what happend, but I just excluded a highly correlated variable. The name did not contain any double points; only alphapetic characters. Anyhow, I'll close it for now.
I am using a logic almost 1:1 copied from R's merge() function to find row match indices over multiple columns. However, they use as column separater "\r", while I am using that cryptic ":", which lead to the error message. I will switch to "\r", which seems to be a better solution.
@inkrement Would be fantastic if you could test your original (failing) example with the main branch
devtools::install_github("ModelOriented/hstats")