Report timing experiment in README

README.Rmd

@@ -296,6 +296,38 @@ list(fit, fit_lav, fit_blav, fit_blavvb) |>
   purrr::map_dbl(\(x) x@timing$total) 
 ```
 
+A little experiment to see how sample size affects run time:
+
+```{r}
+#| fig-width: 8
+#| fig-height: 5
+#| echo: false
+#| warning: false
+#| message: false
+load("inst/timing.RData")
+res |>
+  unnest(c(res_blav, res_blavvb, res_inla), names_sep = "_") |>
+  select(n, ends_with("time")) |>
+  rename(
+    blavaan = res_blav_time,
+    blavaan_vb = res_blavvb_time,
+    INLAvaan = res_inla_time
+  ) |>
+  pivot_longer(
+    cols = -n,
+    names_to = "method",
+    values_to = "time"
+  ) |>
+  ggplot(aes(n, time, col = method)) +
+  geom_point() +
+  geom_smooth(se = FALSE) +
+  scale_x_continuous(labels = scales::comma) +
+  theme_bw() +
+  labs(x = "Sample size", y = "Run time (s)",
+       title = "Total run time to fit two factor SEM with varying sample sizes",
+       caption = "For MCMC sampling, 1000 burnin and 2000 samples were obtained.")
+```
+
 ## Outro
 
 ```{r}

README.md

@@ -116,12 +116,12 @@ mod <- "
 dplyr::glimpse(dat)
 #> Rows: 10,000
 #> Columns: 6
-#> $ y1 <dbl> -1.186116189, 0.235382053, 0.001330662, -1.116578957, -0.537370063,…
-#> $ y2 <dbl> -2.34516559, 0.55765819, -0.83882171, -0.77999471, -0.79058597, 1.3…
-#> $ y3 <dbl> -1.5417462, 1.1229198, -0.3525473, -0.9698537, -1.5521626, 2.163050…
-#> $ y4 <dbl> 1.37136308, 0.20452593, -0.19729736, 0.67467441, 1.01417142, 0.6074…
-#> $ y5 <dbl> 0.93313963, 0.26060807, -0.81823192, 1.12356724, 0.36066546, 0.2903…
-#> $ y6 <dbl> 2.22602893, 1.22153810, -0.88099143, 0.84524435, 0.10090842, 0.4738…
+#> $ y1 <dbl> 1.02192841, 0.29113577, -1.47070927, 0.65444305, -0.77219144, -1.17…
+#> $ y2 <dbl> 0.57817049, 0.39498437, -1.42764854, 0.80022760, -0.37544841, -2.34…
+#> $ y3 <dbl> 0.6068068, 0.3128205, -1.1559359, 1.4147742, -0.1831977, -2.5209732…
+#> $ y4 <dbl> 1.07363543, 0.97521562, 0.12599742, -0.73024295, 0.28783171, 1.5970…
+#> $ y5 <dbl> 1.30085521, 0.62574636, -0.53820263, -0.99175403, 0.80177352, 0.822…
+#> $ y6 <dbl> 1.2908936, 0.7196049, 0.6252699, -1.0316967, 1.3074426, 2.0141520, …
 ```
 
 To fit this model using `{INLAvaan}`, use the familiar `{lavaan}`
@@ -196,6 +196,10 @@ as variational Bayes):
     #>   INLAvaan     lavaan    blavaan blavaan_vb 
     #>     27.785      0.032     47.276     90.578
 
+A little experiment to see how sample size affects run time:
+
+<img src="man/figures/README-unnamed-chunk-6-1.png" width="100%" />
+
 ## Outro
 
 ``` r
@@ -289,6 +293,7 @@ sessioninfo::session_info()
 #>  MASS           7.3-60.2    2024-04-24 [1] local
 #>  Matrix         1.7-0       2024-03-22 [1] CRAN (R 4.4.0)
 #>  matrixStats    1.3.0       2024-04-11 [1] CRAN (R 4.4.0)
+#>  mgcv           1.9-1       2023-12-21 [1] CRAN (R 4.4.0)
 #>  mi             1.1         2022-06-06 [1] CRAN (R 4.4.0)
 #>  minqa          1.2.7       2024-05-20 [1] CRAN (R 4.4.0)
 #>  mnormt         2.1.1       2022-09-26 [1] CRAN (R 4.4.0)

inst/demo.R

@@ -31,11 +31,6 @@ mod <- "
 "
 fit <- isem(model = mod, data = dat, meanstructure = FALSE, verbose = TRUE)
 
-
-
-tmp <- fit
-do.call("inlav_model_test", tmp)
-
 # Political democracy SEM example ----------------------------------------------
 myModel <- '
   # latent variables

inst/timing.R

@@ -74,7 +74,7 @@ run_inlav_sim <- function(N) {
 
 # Run simulations --------------------------------------------------------------
 res <-
-  tibble(n = rep(c(100, 250, 500, 1000, 2000), each = 10)) |>
+  tibble(n = rep(c(100, 250, 500, 1000, 2500, 5000, 7500, 10000), each = 10)) |>
   mutate(
     res_blav = future_map(
       .x = n,
@@ -94,6 +94,7 @@ res <-
       .progress = TRUE
     ),
   )
+save(res, file = "inst/timing.RData")
 
 res |>
   unnest(c(res_blav, res_blavvb, res_inla), names_sep = "_") |>