fixed Delta se

DESCRIPTION

@@ -11,7 +11,7 @@ Description: What the package does (one paragraph).
 License: MIT + file LICENSE
 Encoding: UTF-8
 Roxygen: list(markdown = TRUE)
-RoxygenNote: 7.3.1
+RoxygenNote: 7.3.2
 Imports: 
     blavaan,
     cSEM,

NOTES.md

@@ -1,5 +1,13 @@
 # Notes
 
+MATTHIAS:
+
+- add a warning to the output under "Effect sizes" in case that the indirect effect is larger than the total effect
+- that care should be taken in interpreting the effect size estimates
+- add such a warning to the plot_effect() function as well
+
+- fix website
+
 ## Improvements
 
 - catch misspecifications by the user, for example:
@@ -31,6 +39,14 @@ med_results <- rmedsem(est.full.lpa.mod4, indep="Attractive",
 print(med_results)
 ```
 
+## Bootstrap vs. Monte-Carlo
+
+- implement bootstrap also for CB-SEM
+- implement monte carlo also for cSEM
+- allow options for both methods in the same function
+- also add bias-corrected bootstrap and accelerated bias-corrected bootstrap (see Preacher & Selig, 2012)
+
+
 ## Standard errors
 
 - Stata uses OIM (observed information matrix) for estimating standard errors by default

R/plotting.R

@@ -18,8 +18,8 @@ plot_effect <- function(res, description=TRUE){
   descr.label <- ""
   if(description){
     descr.label <- sprintf("Total effect=%.3f\nThat means %.1f%% of the total effect of '%s' on '%s' is mediated by '%s'.",
-                           es$RIT$tot_eff, 100*es$RIT$ind_eff/es$RIT$tot_eff, res$vars$indep,
-                           res$vars$dep, res$vars$med)
+                           es$RIT$tot_eff, 100*es$RIT$ind_eff/es$RIT$tot_eff,
+                           res$vars$indep, res$vars$dep, res$vars$med)
   }
 
   ggplot2::ggplot(effs, ggplot2::aes(x="", y=value, fill=eff)) +

R/rmedsem_csem.R

@@ -65,7 +65,8 @@ rmedsem.cSEMResults <- function(mod, indep, med, dep,
   sobel_lci <- prodterm - 1.959964*sobel_se
   sobel_uci <- prodterm + 1.959964*sobel_se
 
-  delta_se <- sqrt( (coef_dom^2)*var_moi + (coef_moi^2)*var_dom + (var_moi*var_dom) )
+  #delta_se <- sqrt( (coef_dom^2)*var_moi + (coef_moi^2)*var_dom + (var_moi*var_dom) )
+  delta_se <- sqrt( (coef_dom^2)*var_moi + (coef_moi^2)*var_dom + 2*coef_dom*coef_moi*corrmoidom )
   delta_z  <- prodterm/delta_se
   delta_pv  <- 2*(1-stats::pnorm(abs(delta_z)))
   delta_lci <- prodterm - 1.959964*delta_se

R/rmedsem_lavaan.R

@@ -88,7 +88,9 @@ rmedsem.lavaan <- function(mod, indep, med, dep, standardized=TRUE, mcreps=NULL,
   # in the DELTA METHOD below, it seems like normaly theory limits
   # are used there as well, that is 1.959964 is used
 
-  delta_se <- sqrt( (coef_dom^2)*var_moi + (coef_moi^2)*var_dom + (var_moi*var_dom) )
+  #delta_se <- sqrt( (coef_dom^2)*var_moi + (coef_moi^2)*var_dom + (var_moi*var_dom) )
+  delta_se <- sqrt( (coef_dom^2)*var_moi + (coef_moi^2)*var_dom + 2*coef_dom*coef_moi*corrmoidom )
+
   delta_z  <- prodterm/delta_se
   delta_pv  <- 2*(1-stats::pnorm(abs(delta_z)))
   delta_lci <- prodterm - 1.959964*delta_se