6.3 Make Predictions

• Rules to Live by:
  • Returns a tibble
  • Column names are … erh Predictable
  • Return the same number of rows as are in the data set
    • some predict functions omit observations with NA values. Which is great if that’s what you intend, but if you aren’t expecting that behavior you would have to find out the hard way.

# create example test set
ames_test_small <- ames_test %>% slice(1:5)

# predict on test set
predict(lm_form_fit, new_data = ames_test_small) %>% 
  knitr::kable()

.pred
5.223697
5.221967
5.284242
5.239113
5.314692

• Combining bind_cols with our predict function we can merge our predictions back to the test set.

# add predictions together with actuals
ames_test_small %>%
  select(Sale_Price) %>%
  bind_cols(predict(lm_form_fit, ames_test_small)) %>%
  # Add 95% prediction intervals to the results:
  bind_cols(predict(lm_form_fit, ames_test_small, type = "pred_int")) %>% 
  knitr::kable()

Sale_Price	.pred	.pred_lower	.pred_upper
5.021189	5.223697	4.907039	5.540355
5.235528	5.221967	4.905310	5.538625
5.278525	5.284242	4.967563	5.600922
5.060698	5.239113	4.922441	5.555785
5.596808	5.314692	4.997995	5.631388