offsets do not work correctly in stan_gamm4

Summary:

The offsets don't work right in stan_gamm4. The example here is for a gaussian family, but it is particularly harmful in poisson regression.

Description:

Specifying an offset inside the formula or as argument to the function does not work if the offset referred to is a column inside the data frame. This is also problematic when predicting.

Reproducible Steps:

## example from gamm
library(rstanarm)

dat <- mgcv::gamSim(1, n = 400, scale = 2) ## simulate 4 term additive truth
#> Gu & Wahba 4 term additive model
## Now add 20 level random effect `fac'...
dat$fac <- fac <- as.factor(sample(1:20, 400, replace = TRUE))
dat$y <- dat$y + model.matrix(~ fac - 1) %*% rnorm(20) * .5

br <- stan_gamm4(y ~ s(x0) + x1 + s(x2) + offset(f2), data = dat, random = ~ (1 | fac),
                 chains = 1, iter = 200) # for example speed

RStanARM Version:

2.17.2

R Version:

3.4.3

> sessionInfo()
R version 3.4.3 (2017-11-30)
Platform: x86_64-apple-darwin15.6.0 (64-bit)
Running under: macOS High Sierra 10.13.3

Matrix products: default
BLAS: /Library/Frameworks/R.framework/Versions/3.4/Resources/lib/libRblas.0.dylib
LAPACK: /Library/Frameworks/R.framework/Versions/3.4/Resources/lib/libRlapack.dylib

locale:
[1] de_CH.UTF-8/de_CH.UTF-8/de_CH.UTF-8/C/de_CH.UTF-8/de_CH.UTF-8

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
[1] rstanarm_2.17.2 Rcpp_0.12.14   

loaded via a namespace (and not attached):
 [1] lattice_0.20-35    zoo_1.8-0          gtools_3.5.0       assertthat_0.2.0  
 [5] digest_0.6.12      mime_0.5           R6_2.2.2           plyr_1.8.4        
 [9] stats4_3.4.3       ggplot2_2.2.1      colourpicker_0.3   rlang_0.1.2       
[13] lazyeval_0.2.0     minqa_1.2.4        miniUI_0.1.1       nloptr_1.0.4      
[17] Matrix_1.2-12      DT_0.2             shinythemes_1.1.1  splines_3.4.3     
[21] shinyjs_0.9.1      lme4_1.1-13        stringr_1.2.0      htmlwidgets_0.9   
[25] loo_1.1.0          igraph_1.1.2       munsell_0.4.3      shiny_1.0.5       
[29] compiler_3.4.3     httpuv_1.3.5       rstan_2.17.3       pkgconfig_2.0.1   
[33] base64enc_0.1-3    mgcv_1.8-22        rstantools_1.3.0   htmltools_0.3.6   
[37] tibble_1.3.4       gridExtra_2.3      codetools_0.2-15   threejs_0.3.1     
[41] matrixStats_0.52.2 dplyr_0.7.3        MASS_7.3-47        grid_3.4.3        
[45] nlme_3.1-131       xtable_1.8-2       gtable_0.2.0       magrittr_1.5      
[49] StanHeaders_2.17.2 scales_0.5.0       stringi_1.1.5      reshape2_1.4.2    
[53] bindrcpp_0.2       dygraphs_1.1.1.4   xts_0.10-0         tools_3.4.3       
[57] glue_1.1.1         shinystan_2.4.0    markdown_0.8       crosstalk_1.0.0   
[61] rsconnect_0.8.5    parallel_3.4.3     survival_2.41-3    inline_0.3.14     
[65] colorspace_1.3-2   bayesplot_1.4.0    bindr_0.1         

Operating System:

Your operating system (e.g., OS X 10.11.3)

Start at the stan_gamm4 entry point and run the reproducible example from the issue. Compare offsets supplied inside the formula or as a function argument when the offset is a data-frame column, including prediction behavior. Done means column-based offsets work correctly for fitting and prediction, particularly for poisson regression.