map_models: Create multiple models which vary by one parameter

In rmgpanw/ukbwranglrextra: Exploring UKB Data - Extra

Create multiple models which vary by one parameter

Description

Use when creating multiple models which vary by one parameter. See example below.

Usage

map_models(
  df,
  params,
  model_str,
  engine = parsnip::set_engine(object = parsnip::logistic_reg(), engine = "glm"),
  rm_raw_model = TRUE
)

Arguments

df	Data frame.
params	A character vector of values to substitute .x in model_str.
model_str	A model formula as a string, containing one variable as .x.
engine	A parsnip engine.
rm_raw_model	If TRUE (the default), the raw models are removed.

Details

Ideally aim to supply a df without missing data. Otherwise, check N observations for models (in fit column, model_glance) to see if many (or any) observations were removed.

Value

A data frame

Examples

# dummy dataset
mtcars2 <- mtcars %>%
 tibble::rownames_to_column(var = "make") %>%
 dplyr::mutate("is_merc" = ifelse(stringr::str_detect(.data[["make"]], pattern = "^Merc"),
                                  yes = 1,
                                  no = 0 )) %>%
 tibble::as_tibble() %>%
 dplyr::mutate(dplyr::across(tidyselect::starts_with("is_"), as.factor))

# preview dummy dataset
mtcars2

# run `map_models()` - logistic regression to predict Mercedes make
result <-
map_models(
  df = mtcars2,
  model_str = "is_merc ~ mpg + cyl + disp + {.x}",
  params = c(GEAR = 'gear', CARB = 'carb'),
  engine = parsnip::set_engine(object = parsnip::logistic_reg(), engine = "glm"),
  rm_raw_model = FALSE
)

# view result
result

# model outputs are stored in the `fit` column
names(result$fit[[1]])

# 'tidy' model outputs are under `model_tidy`
result$fit[[1]]$model_tidy

# 'glance' model outputs are under `model_glance`
result$fit[[1]]$model_glance

# `model_raw` contains either the raw model under `result`, or an error message under `error`
names(result$fit[[1]]$model_raw)

rmgpanw/ukbwranglrextra documentation built on Jan. 4, 2023, 1:56 a.m.