store_boot(type="std.all") / standardizedSolution_boot() can fail with dimnames length mismatch when std_names() drops rows via merge()

[dgerdesmann]

Hi, thanks for making semboottools. I ran into a bootstrapping failure for standardized coefficients that seems to be caused by std_names(). Calling store_boot(..., type = "std.all") and then standardizedSolution_boot() (or any call path that triggers boot_est_std()) failed in a few (but not all) more complex models with:

Error in dimnames(x) <- dn :
  length of 'dimnames' [2] not equal to array extent
Called from: `colnames<-`(`*tmp*`, value = std_names(object))

The error happens when boot_est_std() builds a bootstrap matrix out_all and then does:

colnames(out_all) <- std_names(object)

So if merge(std, ptable, all.y = FALSE) in std_names() returns a data frame with a different number of rows than std, std_names() returns the wrong-length name vector and the colnames<- assignment errors.

I patched this by avoiding merge and relying on if-else label name/finding, so different lengths can't occur:

std_names_nomerge <- function(object, type = "std.all", ...) {
  std <- lavaan::standardizedSolution(
    object, type = type, se = FALSE, output = "data.frame", ...
  )

  lab <- if ("label" %in% names(std)) std$label else rep("", nrow(std))

  out <- ifelse(
    std$op == ":=",
    std$lhs,
    ifelse(!is.na(lab) & lab != "", lab,
           paste(std$lhs, std$op, std$rhs))
  )

  make.unique(out)
}

assignInNamespace("std_names", std_names_nomerge, ns = "semboottools")

Worked for me, but might not be great as a general approach.

Couldn't figure out a minimal repex, so I just described the issue. Hope this helps.

[sfcheung]

Thanks for reporting this issue. We tried some scenarios, but could not yet produce the problem. A lot of types of models are supported by lavaan, and there may be some we haven't considered.

May we learn more about the nature of the models for which you encountered that error? Thanks in advance.