missing = "pairwise" does not take effect

[psychelzh]

For example,

dat <- matrix(rnorm(1000), ncol = 10)
dat[sample(length(dat), 100)] <- NA
x <- kfa::kfa(dat, missing = "pairwise")
#> Power analysis indicates the sample size is too small for k-fold cross validation.
#>     Adjust assumptions or manually create a single holdout sample.
#> [1] "Using 19 cores for parallelization."
#> [1] "Finished EFAs. Starting CFAs"
lavaan::lavInspect(x$cfas[[1]][[1]], "options")$missing
#> [1] "listwise"

^{Created on 2024-04-14 with reprex v2.1.0}

[psychelzh]

Based on these lines:

kfa/R/k_cfa.R

Lines 16 to 21 in a91d785

	sampstats <- sample_stats(data = data,
	variables = variables,
	ordered = ordered,
	estimator = estimator,
	missing = missing,
	...)

and these lines:

kfa/R/k_cfa.R

Lines 31 to 41 in a91d785

	fit <- lavaan::cfa(model = syntax[[c]],
	sample.cov = sampstats$cov,
	sample.nobs = sampstats$nobs,
	sample.th = sampstats$th,
	# sample.mean = sampstats$mean,
	meanstructure = FALSE,
	WLS.V = sampstats$wls.v,
	NACOV = sampstats$nacov,
	estimator = estimator,
	missing = missing,
	parameterization = "delta")

I understand it now. pairwise is used in sample_stats(), from which the statistics are calculated. And these statistics are then fed to cfa() and that, probably, lavaan will ignore the missing parameter then. Correct me in case I am wrong.

[knickodem]

You are correct. Nonetheless, I am going to keep this open as a reminder to add documentation and testing regarding this point.