update vignettes

R/mediation.R

@@ -262,7 +262,7 @@ from_summarized_experiment <- function(exper, outcomes, treatments, mediators,
 #'   Defaults to NULL, in which case the pretreatments slot will remain empty.
 #' @return result An object of class `mediation_data`, with separate slots for
 #'   each of the node types in a mediation analysis diagram.
-#' @importFrom phyloseq otu_table
+#' @importFrom phyloseq otu_table sample_data
 #' @noRd
 from_phyloseq <- function(exper, outcomes, treatments, mediators, pretreatments) {
   if (attr(otu_table(exper), "taxa_are_rows")) {
@@ -327,6 +327,25 @@ from_data_frame <- function(df, outcomes, treatments, mediators,
 #' @return An object of class `multimedia` encapsulating the full mediation
 #'   model and data.
 #' @seealso multimedia-class
+#' @examples
+#' exper <- demo_joy() |>
+#'   mediation_data("PHQ", "treatment", starts_with("ASV"))
+#' multimedia(exper)
+#' 
+#' exper <- demo_spline(tau = c(2, 1)) |>
+#'   mediation_data(starts_with("outcome"), "treatment", "mediator")
+#' multimedia(exper)
+#' 
+#' # real data example with a pretreatment variable
+#' data(mindfulness)
+#' exper <- mediation_data(
+#'  mindfulness,
+#'  phyloseq::taxa_names(mindfulness),
+#'  "treatment",
+#'  starts_with("mediator"),
+#'  "subject"
+#' )
+#' multimedia(exper)
 #' @export
 multimedia <- function(mediation_data,
                        outcome_estimator = lm_model(),

vignettes/IBD.Rmd

@@ -410,8 +410,10 @@ plot_mediators(indirect_sorted, samples, treatment = "Study.Group")
 ```
 
 Finally, we can look at direct effects from this model, which can be compared to
-the analogous figure for the sparse regression. Note that now, many outcomes
-with exact zeros under treatment have been detected.
+the analogous figure for the sparse regression. Now, many outcomes with exact
+zeros under treatment have been detected, which is consistent with the the
+hurdle model's effectiveness in accurately reflecting features that influence
+zero-inflation.
 
 ```{r hurdle_direct, fig.width = 6, height = 5}
 direct <- direct_effect(model, exper) |>

vignettes/illustration.Rmd

@@ -42,7 +42,6 @@ th <- theme_classic() +
     legend.box.background = element_rect(fill = "transparent"),
     legend.position = "bottom"
   )
-
 theme_set(th)
 ```
 

vignettes/mindfulness.Rmd

@@ -43,14 +43,13 @@ th <- theme_classic() +
     legend.box.background = element_rect(fill = "transparent"),
     legend.position = "bottom"
   )
-
 theme_set(th)
 ```
 
 ```{r helpers, include = FALSE}
 normalize <- function(x) {
   x <- as_tibble(x)
-  x / rowSums(x)
+  clr(x / rowSums(x))
 }
 ```
 
@@ -69,7 +68,6 @@ other of these possibilities.
 library(compositions)
 library(ggdist)
 library(ggrepel)
-library(patchwork)
 library(phyloseq)
 library(tidyverse)
 library(multimedia)
@@ -262,7 +260,6 @@ ggplot(combined_samples) +
   labs(x = "log(Relative Abundance)", y = "Genus", fill = "", col = "") +
   facet_grid(~source, scales = "free") +
   theme(panel.border = element_rect(fill = NA))
-ggsave("~/Downloads/lnm_samples.png", dpi = 400)
 ```
 
 

vignettes/random.Rmd

@@ -43,7 +43,6 @@ th <- theme_classic() +
     legend.box.background = element_rect(fill = "transparent"),
     legend.position = "bottom"
   )
-
 theme_set(th)
 ```
 
@@ -168,12 +167,11 @@ indirect_pathwise(model, exper)
 
 Effect estimates are rarely enough on their own. We need some uncertainty
 assessments to set appropriate expectations. The most straightforward approach
-is to use the bootstrap.  Each element in `funs` will get its own data.frame
-with the bootstrap distribution for that estimator.
+is to use the bootstrap. Each function in the third argument, `fs`, will get its
+own data.frame with the bootstrap distribution for that estimator.
 
 ```{r bootstrap}
-funs <- list(direct_effect = direct_effect)
-bootstrap(model, exper, funs, B = 100)
+bootstrap(model, exper, c(direct_effect = direct_effect))
 ```
 
 We can also generate synthetic nulls to calibrate selection sets. The third