In awconway/eicu_nitro: What the Package Does (One Line, Title Case)

---

title: "Data for models" output: html_document

---

library(targets)
library(tidyverse)

These targets are used to filter out observations where the bp was measured more than 15 minutes before and after a dose change

```{targets timeBefore} tar_target(time_before, 15)

```{targets timeAfter}
  tar_target(time_after, 15)

First we filter to the data to we need and create splits for cross-fold validation.

```{targets multiples, tar_simple=TRUE} dataModel |> group_by (id) |> count() |> arrange(desc(n)) |> filter(n > 1) |> pull (id)

```{targets trainingResponse, tar_simple=TRUE}
 dataModel |> 
    filter(id %in% multiples) |>
    group_by(id) |>
    mutate(lag_nitro_pre = lag(nitro_pre),
           lag_nitro_post = lag(nitro_post),
            lag_sbp_pre = lag(sbp_pre),
           lag_sbp_post = lag(sbp_post),
           lag_nitro_diff = lag_nitro_post - lag_nitro_pre,
           lag_sbp_diff = lag_sbp_post - lag_sbp_pre) |> 
    mutate(first_titration = row_number(n_nitro)) |>
    filter(first_titration != 1) |> 
  ungroup()

```{targets foldsIndexResponse} tar_target(foldsIndexResponse, make_kfold(trainingResponse, 5))

```{targets foldsFiveResponse}
  tar_target(foldsFiveResponse, 
  group_vfold_cv(foldsIndexResponse,
    group = fold,
    v = 5
  ))

Next we find the baseline metric for comparison. The predictions need to be better than just passing in the 'pre' sbp as a prediction - we'll call this the baseline.

{targets baselineTrainingResponse} tar_target(baselineTrainingResponse, trainingResponse |> metrics( truth = sbp_post, estimate = sbp_pre ))

awconway/eicu_nitro documentation built on Feb. 21, 2022, 4:35 p.m.