Guys,
Hello!
I am testing different Stan models setups with Stan models and a rstanarm model.
I am seeing a lot of differences in computation times between Stan and rstanarm. Even with the QR reparameterization with Stan.
I’ve simulated data with the following dimensions
X: 10.000 obs and 7 predictors (Matrix)
y: a linar combination of the Xs (Vector)
Here’s the setup:
Stan
-
stan.stan: Simple multivariate gaussian linear model -
stan_std.stan: Same asstan.stanwith the data standardized to mean 0 and variance/sd 1 -
stan_qr.stan: Same asstan.stanwith the QR decomposition (to speed up things)
rstanarm
-
stan_glm(y ~ .): default parameterQR = TRUE
In all samplings, I am using the default arguments (4 chains, 2000 iterations) and I have in .Rprofile file the multicore option enabled options(mc.cores = parallel::detectCores()).
For all 3 Stan models I am seeing a run time of around ~900s (with the QR a little more fast than normal and std)
But for rstanarm I am seeing a run time of 13s!!!
What am I doing wrong? I’ve pre-compiled all models before I’ve run them.
This is my session info in a 2018 MacBook Pro 13" i7 2.7 GHz Quad-Core Intel Core i7 16GB RAM with 321GB SSD free
R version 4.0.2 (2020-06-22)
Platform: x86_64-apple-darwin17.0 (64-bit)
Running under: macOS Catalina 10.15.6
Matrix products: default
BLAS: /System/Library/Frameworks/Accelerate.framework/Versions/A/Frameworks/vecLib.framework/Versions/A/libBLAS.dylib
LAPACK: /Library/Frameworks/R.framework/Versions/4.0/Resources/lib/libRlapack.dylib
locale:
[1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8
attached base packages:
[1] stats graphics grDevices utils datasets methods base
other attached packages:
[1] microbenchmark_1.4-7 rstan_2.21.2 ggplot2_3.3.2 StanHeaders_2.21.0-5 rstanarm_2.21.1
[6] Rcpp_1.0.5
loaded via a namespace (and not attached):
[1] splines_4.0.2 jsonlite_1.7.0 gtools_3.8.2 RcppParallel_5.0.2 threejs_0.3.3 shiny_1.5.0
[7] assertthat_0.2.1 statmod_1.4.34 stats4_4.0.2 yaml_2.2.1 pillar_1.4.6 lattice_0.20-41
[13] glue_1.4.1 digest_0.6.25 promises_1.1.1 minqa_1.2.4 colorspace_1.4-1 Matrix_1.2-18
[19] htmltools_0.5.0 httpuv_1.5.4 plyr_1.8.6 dygraphs_1.1.1.6 pkgconfig_2.0.3 purrr_0.3.4
[25] xtable_1.8-4 scales_1.1.1 processx_3.4.3 later_1.1.0.1 lme4_1.1-23 tibble_3.0.3
[31] bayesplot_1.7.2 generics_0.0.2 ellipsis_0.3.1 DT_0.15 withr_2.2.0 shinyjs_1.1
[37] cli_2.0.2 survival_3.2-3 magrittr_1.5 crayon_1.3.4 mime_0.9 ps_1.3.3
[43] fansi_0.4.1 nlme_3.1-148 MASS_7.3-51.6 xts_0.12-0 pkgbuild_1.1.0 colourpicker_1.0
[49] rsconnect_0.8.16 tools_4.0.2 loo_2.3.1 prettyunits_1.1.1 lifecycle_0.2.0 matrixStats_0.56.0
[55] stringr_1.4.0 V8_3.2.0 munsell_0.5.0 callr_3.4.3 compiler_4.0.2 rlang_0.4.7
[61] nloptr_1.2.2.2 grid_4.0.2 ggridges_0.5.2 rstudioapi_0.11 htmlwidgets_1.5.1 crosstalk_1.1.0.1
[67] igraph_1.2.5 miniUI_0.1.1.1 base64enc_0.1-3 boot_1.3-25 gtable_0.3.0 codetools_0.2-16
[73] inline_0.3.15 curl_4.3 markdown_1.1 reshape2_1.4.4 R6_2.4.1 gridExtra_2.3
[79] rstantools_2.1.1 zoo_1.8-8 dplyr_1.0.1 fastmap_1.0.1 shinystan_2.5.0 shinythemes_1.1.2
[85] stringi_1.4.6 parallel_4.0.2 vctrs_0.3.2 tidyselect_1.1.0
rstan-vs-rstarnarm.R (947 Bytes) stan_qr.stan (836 Bytes) stan_std.stan (684 Bytes) stan.stan (516 Bytes)