[R-meta] Non positive V from metafor::rcalc()

[Yuhang Hu]

Hello All,

In my data below, I have numerous NAs for the correlations reported for a
set of 8 variables of interest across my studies. Two questions:

1- rma.mv() says that my V is non-positive definite. To what extent can I
ignore this and if not ignorable, what other options do I have?

2- How can I convert back the 'coef(res)' and 'vcov(res)' to the original
"r" scale?

Thanks,
Yuhang

# Data and code: #

dat <- read.csv("https://raw.githubusercontent.com/ilzl/i/master/j.csv")

tmp <- rcalc(ri ~ var1 + var2 | Study, ni=N, data=dat, rtoz = TRUE)
V <- tmp$V
dat <- tmp$dat

res <- rma.mv(yi~ var1.var2 - 1, V,
               random = ~var1.var2 | Study,
               data=dat, control = list(nearpd=TRUE), sparse = TRUE) |>
  robust(cluster = Study, clubSandwich = TRUE)

	[[alternative HTML version deleted]]