Linear regression with tidymodels

Modeling with tidymodels in R

David Svancer

Data Scientist

Model fitting with parsnip

Linear regression model

Predicting hwy using cty as a predictor

$$hwy = \beta_{0} + \beta_{1} cty$$

Model parameters

$ \beta_{0} $ is the intercept
$ \beta_{1} $ is the slope

Highway versus city fuel efficiency

Linear regression model

Predicting hwy using cty as a predictor

$$hwy = \beta_{0} + \beta_{1} cty$$

Model parameters

$ \beta_{0} $ is the intercept
$ \beta_{1} $ is the slope

Estimated paramters from training data

$$\small hwy = 0.77 + 1.35(cty)$$

Mpg data with linear regression line

Model formulas

Model formulas in parsnip

Used to assign column roles
- Outcome variable
- Predictor variables

General form

outcome ~ predictor_1 + predictor_2 + ...

Shorthand notation

outcome ~ .

Predicting hwy using cty as a predictor variable

hwy ~ cty

The parsnip package

Unified syntax for model specification in R

Specify the model type
- Linear regression or other model type
Specify the engine
- Different engines correspond to different underlying R packages
Specify the mode
- Either regression or classification

Parsnip package

Fitting a linear regression model

Define model specification with parsnip

linear_reg()

Pass lm_model to the fit() function

Specify model formula
data to use for model fitting

lm_model <- linear_reg() %>%

  set_engine('lm') %>%

  set_mode('regression')

lm_fit <- lm_model %>% 
  fit(hwy ~ cty, data = mpg_training)

Obtaining the estimated parameters

The tidy() function

Takes a trained parsnip model object
Creates a model summary tibble
term and estimate column provide estimated parameters

tidy(lm_fit)

# A tibble: 2 x 5
  term        estimate std.error statistic  p.value
  <chr>          <dbl>     <dbl>     <dbl>    <dbl>
1 (Intercept)    0.769    0.528       1.46 1.47e- 1
2 cty            1.35     0.0305     44.2  6.32e-97

Making predictions

Pass trained parsnip model to the predict() function

new_data specifies dataset on which to predict new values

Standardized output from predict()

Returns a tibble
Keep rows in the same order as new_data input
Names prediction column .pred

hwy_predictions <- lm_fit %>% 
  predict(new_data = mpg_test)


hwy_predictions

# A tibble: 57 x 1
   .pred
   <dbl>
 1  25.0
 2  27.7
 3  25.0
 4  25.0
 5  22.3
# ... with 47 more rows

Adding predictions to the test data

The bind_cols() function

Combines two or more tibbles along the column axis
Useful for creating a model results tibble

Steps

Select hwy and cty from mpg_test
Pass to bind_cols() and add predictions column

mpg_test_results <- mpg_test %>%
  select(hwy, cty) %>%

  bind_cols(hwy_predictions)

mpg_test_results

# A tibble: 57 x 3
     hwy   cty .pred
   <int> <int> <dbl>
 1    29    18  25.0
 2    31    20  27.7
 3    27    18  25.0
 4    26    18  25.0
 5    25    16  22.3
# ... with 47 more rows

Let's model!

Modeling with tidymodels in R