Open Ojami opened 1 year ago
Hi @Ojami
I don't know that I've seen the p-values since there are three indirect effects. Most people use the confidence interval from the bootstrap to determine if it includes zero or not. I suppose you could calculate sobel separately for each indirect effect, but I'm not sure how you'd do it for the two part mediation. Do you have a reference I could look at?
erin
Hi Erin,
Thanks for your response. Yes, I've seen mostly bootstrapped CIs only. However, lavaan::sem()
outputs z
and p
(an example).
Also, if I'm not wrong mediation1
uses Sobel for z
, while mediation::mediate()
uses bootstrapped P directly (boot$t
):
pval <- function(x, xhat){
## Compute p-values
if (xhat == 0) out <- 1
else {
out <- 2 * min(sum(x > 0), sum(x < 0)) / length(x)
}
return(min(out, 1))
}
So, I was wondering why one cannot use the same approach for serial mediation analysis? Given a large sample size, I suppose (?) bootstrap Z (boot.results$t0/sd(boot.results$t))
should converge to Sobel's Z. Am I mistaken?
Best/Oveis
Good points. I'll check that out. And I would think so ... the t-values for all of them are in there saved under indirect, so then you could calculate p with pt(T_VALUE, df = DF_VALUE, lower.tail = FALSE)
.
erin
Hi and thanks for this great package,
I was wondering how one can calculate P for the indirect effect in the
mediation2
func?Best/Oveis