model blueprints

blueprint.Rmd

## Create a model blueprint, using a consistent API across modeling platforms.

```{r}
blueprint <- make_blueprint("nonmem")
```

```{r}
one_cmt <- blueprint %>% 
  ode_model(cmt = 1, depot = 1) %>%
  parameterize("CL") %>% 
  params(CL = 1.6, V = 14.3, KA = 0.8)
```

```{r}
one_cmt_block <- one_cmt  %>%
  heirarchy(block(CL = 0.1, V = 0.04, CL_V = 0.01), KA = 0.1) # add a block omega for Cl/V and a separate omega for KA

one_cmt_diag <- one_cmt  %>%
  heirarchy(CL = 0.1, V = 0.04, KA = 0.1) # add diagonal matrix for Cl, V, KA
```

At this point, we may want to try different residual error structures

```{r}
one_cmt_block_combined <- one_cmt_block %>% residual_error(add = 1, prop=0.04)
one_cmt_block_prop <- one_cmt_block %>% residual_error(prop=0.04)
```

Or both can apply something to multiple models

```{r}
model_list <- c(one_cmt_block, one_cmt_diag)

map(model_list, function(model) {
  model %>% residual_error(prop=0.04)
})
```


This iterative construction is the same as:

```{r}
make_blueprint("mrgsolve") %>% 
  ode_model(cmt = 1, depot = 1) %>%
  parameterize("CL") %>% 
  params(CL = 1.6, V = 14.3, KA = 0.8) %>%
  heirarchy(block(CL = 0.1, V = 0.04, CL_V = 0.01), KA = 0.1) %>% 
  residual_error(prop=0.04)
```

As the arguments can be chained, the later steps can be created as a functional blueprint

```{r}
one_cmt_creator <- . %>% 
  ode_model(cmt = 1, depot = 1) %>%
  parameterize("CL") %>% 
  params(CL = 1.6, V = 14.3, KA = 0.8) %>%
  heirarchy(block(CL = 0.1, V = 0.04, CL_V = 0.01), KA = 0.1) %>% 
  residual_error(prop=0.04)
```


```{r}
make_blueprint("mrgsolve") %>% one_cmt_creator()
make_blueprint("nonmem") %>% one_cmt_creator()
```