further tweaks to vignettes, run-extended

.Rbuildignore

@@ -9,6 +9,7 @@
 ^_pkgdown.yml$
 ^doc$
 ^Meta$
+^R/rebuild_vignettes\.R$
 ^vignettes/.*\.qmd\.orig$
-^vignettes/rebuild_vignettes\.R$
 ^vignettes/*_files$
+^vignettes/*\.R$

DESCRIPTION

@@ -17,7 +17,7 @@ Description: Easy-to-use and efficient interface for
   multiple measurements per individual, time-varying and time-invariant 
   effects, and a wide range of discrete and continuous distributions. 
   Estimation of these dynamic multivariate panel models is carried out via 
-  'Stan'. For an  in-depth tutorial of the package, see 
+  'Stan'. For an in-depth tutorial of the package, see 
   (Tikka and Helske, 2024) <doi:10.48550/arXiv.2302.01607>.
 License: GPL (>= 3)
 URL: https://docs.ropensci.org/dynamite/, https://github.com/ropensci/dynamite/
@@ -49,8 +49,7 @@ Suggests:
     quarto,
     testthat (>= 3.0.0),
     tidyr
-VignetteBuilder: 
-    quarto
+VignetteBuilder: quarto
 Config/testthat/edition: 3
 Encoding: UTF-8
 Roxygen: list(markdown = TRUE, roclets = c ("namespace", "rd",

R/rebuild_vignettes.R

@@ -0,0 +1,16 @@
+# Working directory needs to be /vignettes
+
+rebuild_vignettes <- function() {
+  files <- c(
+    "dynamite_custom",
+    "dynamite_simulation"
+  )
+  for (i in seq_along(files)) {
+    e <- new.env()
+    knitr::knit(
+      input = paste0(files[i], ".qmd.orig"),
+      output = paste0(files[i], ".qmd"),
+      envir = e
+    )
+  }
+}

vignettes/dynamite.qmd

@@ -9,8 +9,6 @@ format:
     toc: true
     number-sections: true
     html-math-method: mathjax
-pkgdown:
-  as_is: true
 bibliography: dynamite.bib
 vignette: >
   %\VignetteIndexEntry{dynamite: An R Package for Dynamic Multivariate Panel Models}
@@ -325,7 +323,6 @@ We can obtain posterior samples or summary statistics of the model using the `as
 #| out.width: "100%"
 #| fig.width: 8
 #| fig.height: 4
-#| fig.dpi: 300
 #| fig.align: "center"
 #| fig.cap: "Posterior means and 90\\% posterior intervals of the time-invariant coefficients for the multivariate model."
 library("ggplot2")
@@ -409,7 +406,6 @@ The resulting data frame `sumrb` is equal to the previous `sumr` (apart from sto
 #| out.width: "100%"
 #| fig.width: 8
 #| fig.height: 3.5
-#| fig.dpi: 300
 #| fig.align: "center"
 #| fig.cap: "Expected causal effects of interventions $do(b_5 = 0)$ and $do(b_5 = 1)$ on $g_t$. The black lines show the posterior means and the gray areas show 90\\% posterior intervals."
 ggplot(sumr, aes(time, mean)) +
@@ -452,7 +448,6 @@ We can also plot the difference between the expected causal effects as shown in
 #| out.width: "100%"
 #| fig.width: 8
 #| fig.height: 3.5
-#| fig.dpi: 300
 #| fig.align: "center"
 #| fig.cap: "Difference between the expected causal effects $E(g_t | do(b_5 = 1)) - E(g_t | do(b_5 = 0))$. The black line shows the posterior mean and the gray area shows a 90\\% posterior interval."
 ggplot(sumr_diff, aes(time, mean)) +
@@ -694,13 +689,11 @@ A `plot()` method is available for `"dynamiteformula"` objects that can be used
 ```{r}
 #| label: fig-multichanneldagplot
 #| fig.show: "hold"
-#| out.width: "50%"
-#| fig.cap: "DAGs for the multivariate model created using the `plot()` method for `\"dynamitefit\"` objects. The left panel shows the model structure including the auxiliary response variable `logp` while the right panel shows the model structure where the auxiliary variable is not included. The latter DAG is obtained via a functional projection where the parents of `logp` become the parents of the children of `logp` and `logp` is removed from the graph at each timepoint."
+#| out.width: "100%"
+#| fig.cap: "DAGs for the multivariate model created using the `plot()` method for `\"dynamitefit\"` objects. Panel (a) shows the model structure including the auxiliary response variable `logp` while panel (b) shows the model structure where the auxiliary variable is not included. The latter DAG is obtained via a functional projection where the parents of `logp` become the parents of the children of `logp` and `logp` is removed from the graph at each timepoint."
 #| fig.subcap:
 #|   - ""
 #|   - ""
-#| layout-ncol: 2
-#| fig.dpi: 300
 plot(multi_formula)
 plot(multi_formula, show_auxiliary = FALSE)
 ```
@@ -877,7 +870,6 @@ For instance, @fig-parameterposteriorplot shows the posterior means and posterio
 #| out.width: "100%"
 #| fig.width: 10
 #| fig.height: 4
-#| fig.dpi: 300
 #| fig.align: "center"
 #| fig.cap: "Posterior means (black lines) and 90\\% posterior intervals (gray areas) for the time-varying coefficients for the response variable `y` in the `gaussian_example_fit` model. The panels from left to right show the time-varying intercept for `y`, the time-varying effect of `x` on `y`, and the time-varying effect of `lag(y)` (the previous time-point) on `y`."
 plot(
@@ -894,7 +886,6 @@ While `plot_type = "default"` produces plots such as @fig-parameterposteriorplot
 #| out.width: "100%"
 #| fig.width: 9
 #| fig.height: 4
-#| fig.dpi: 300
 #| fig.align: "center"
 #| fig.cap: "Marginal posterior density and traceplot of the MCMC chains of the time-invariant regression coefficient `beta_y_z` of `z` for the response variable `y` in the `gaussian_example_fit` model."
 plot(gaussian_example_fit, plot_type = "trace", types = "beta")
@@ -936,7 +927,6 @@ For example, we can obtain posterior predictive samples for the first 4 groups i
 #| out.width: "100%"
 #| fig.width: 8
 #| fig.height: 5
-#| fig.dpi: 300
 #| fig.align: "center"
 #| fig.cap: "Posterior predictive samples for the first 4 groups of the `gaussian_example` data. Lines in red represent the observed values."
 pred <- predict(gaussian_example_fit, n_draws = 50)
@@ -1001,7 +991,7 @@ In the future, we plan to extend the capabilities of **dynamite** by adding supp
 ::: {#refs}
 :::
 
-# Details on latent factors {#sec-latentfactor .appendix}
+# Details on latent factors {#sec-latentfactor .unnumbered}
 
 Latent factor models with product terms $\lambda_i\psi_t$ are known to suffer from identifiability issues. For example, it is possible to multiply each $\lambda_i$ by some constant $c$ while simultaneously multiplying $\psi_i, t=1,\ldots, T$ with the reciprocal of the same constant, leading to the same likelihood value as the original model. In case of multiple latent factors and (vector) autoregressive process on $\psi_t$, @bai2015 discuss two alternative identifiability constraints, which in our single factor model translate to fixing $\lambda_i = 1$ for some $i$, or constraining $\lambda_i > 0$ for some $i$, with an additional constraint that the standard deviation of the noise term of $\psi_t$ is 1. In both cases, we need to decide which individual is used as a reference for the constrained $\lambda_i$. This choice can lead to computational issues if the true value of $\lambda_i$ is not compatible with these constrains (e.g., the true value is zero). Instead, we define the constraints via the mean of $\lambda$.
 

vignettes/dynamite_custom.qmd

@@ -5,8 +5,6 @@ format:
     theme: none
     minimal: true
     embed-resources: true
-pkgdown:
-  as_is: true
 vignette: >
   %\VignetteIndexEntry{Model customization and non-MCMC estimation with dynamite}
   %\VignetteEngine{quarto::html}
@@ -70,7 +68,7 @@ It is important to note that in order to use the post-processing functions of `d
 plot(fit, type = "nu", n_params = 20)
 ```
 
-![](nuplot-1.png)
+<img src="nuplot-1.png" width="100%" />
 
 perform leave-one-out cross-validation:
 
@@ -107,7 +105,7 @@ ggplot2::ggplot(pp, ggplot2::aes(time, y_new, group = .draw)) +
   ggplot2::theme_bw()
 ```
 
-![](custommodelplot-1.png)
+<img src="custommodelplot-1.png" width="100%" />
 
 # Using non-MCMC algorithms in `dynamite`
 
@@ -160,4 +158,4 @@ ggplot2::ggplot(pp2, ggplot2::aes(time, y_new, group = .draw)) +
   ggplot2::theme_bw()
 ```
 
-![](vbplot-1.png)
+<img src="vbplot-1.png" width="100%" />

vignettes/dynamite_custom.qmd.orig

@@ -5,8 +5,6 @@ format:
     theme: none
     minimal: true
     embed-resources: true
-pkgdown:
-  as_is: true
 vignette: >
   %\VignetteIndexEntry{Model customization and non-MCMC estimation with dynamite}
   %\VignetteEngine{quarto::html}
@@ -98,7 +96,7 @@ It is important to note that in order to use the post-processing functions of `d
 ```{r}
 #| label: nuplot
 #| fig.cap: ""
-#| fig.dpi: 300
+#| out.width: "100%"
 plot(fit, type = "nu", n_params = 20)
 ```
 
@@ -111,7 +109,7 @@ and compute predictions:
 ```{r}
 #| label: custommodelplot
 #| fig.cap: ""
-#| fig.dpi: 300
+#| out.width: "100%"
 newdata <- data.frame(
   time = 1:30,
   id = 51,
@@ -172,7 +170,7 @@ summary(fit_vb_dynamite, types = "beta")
 ```{r}
 #| label: vbplot
 #| fig.cap: ""
-#| fig.dpi: 300
+#| out.width: "100%"
 pp2 <- predict(
   fit_vb_dynamite, newdata = newdata, new_levels = "original", n_draws = 50
 )

vignettes/dynamite_priors.qmd

@@ -6,8 +6,6 @@ format:
     minimal: true
     embed-resources: true
     html-math-method: mathjax
-pkgdown:
-  as_is: true
 vignette: >
   %\VignetteIndexEntry{Priors for dynamic multivariate panel models}
   %\VignetteEngine{quarto::html}

vignettes/dynamite_simulation.qmd

@@ -5,8 +5,6 @@ format:
     theme: none
     minimal: true
     embed-resources: true
-pkgdown:
-  as_is: true
 vignette: >
   %\VignetteIndexEntry{Simulating data from a dynamic multivariate panel model}
   %\VignetteEngine{quarto::html}
@@ -181,6 +179,7 @@ Again, we apply the `get_parameter_dims()` function to get the model parameters
 
 ``` r
 get_parameter_dims(x = f, data = d, time = "time", group = "id")
+#> recompiling to avoid crashing R session
 #> $omega_y
 #> [1]  2 10
 #> 

vignettes/dynamite_simulation.qmd.orig

@@ -5,8 +5,6 @@ format:
     theme: none
     minimal: true
     embed-resources: true
-pkgdown:
-  as_is: true
 vignette: >
   %\VignetteIndexEntry{Simulating data from a dynamic multivariate panel model}
   %\VignetteEngine{quarto::html}
@@ -249,7 +247,6 @@ We can plot individual trajectories of `y` over time to visualize the data.
 #| fig.width: 7
 #| fig.height: 3.5
 #| fig.cap: ""
-#| fig.dpi: 300
 #| out.width: "100%"
 library("ggplot2")
 sim |>