Ex 4 on Lab 5

[erincperry]

@LucyMcGowan I am unable to get my code for the exercise 4 to run This is what I have so far:

set.seed(7)
Wage_cv <- vfold_cv(Wage, v = 10)
results <- fit_resamples(linear_spec,
              preprocessor = wage_rec,
              resamples = Wage_cv)
grid <- expand_grid(deg_free = seq(1,10, by = 1)) 
tuning <- tune_grid(linear_spec,
                   wage_rec,
                    grid = grid,
                   resamples = Wage_cv)

tuning %>%
 collect_metrics() %>%
  filter(.metric == "deg_free") %>%
  arrange(mean())

The error says x Fold01: recipe: Error in df - 1L: non-numeric argument to ... x Fold02: recipe: Error in df - 1L: non-numeric argument to ... x Fold03: recipe: Error in df - 1L: non-numeric argument to ...

Do you have any advice / suggestions on how to fix this?

[LucyMcGowan]

Can you send your recipe code?

[erincperry]

wage_rec <- recipe(wage ~ age, health_ins, jobclass, education, race, data = Wage) %>% step_ns(age, deg_free = tune()) wage_rec

[LucyMcGowan]

Looks like a type (, should be +)

wage_rec <- recipe(wage ~ age + health_ins + jobclass + education +race, data = Wage) %>%
step_ns(age, deg_free = tune())
wage_rec

[erincperry]

Thank you! I changed to the plus signs, but I am still getting the same error when I run the last code. Is there anything else I should try?

[LucyMcGowan]

Ahh I see it 👀 the problem is actually here:

tuning %>%
 collect_metrics() %>%
  filter(.metric == "deg_free") %>%
  arrange(mean())

the filter line is filtering the .metric column. This column has two potential values rmse or rsq, you want to use rmse so that is what you should filter based on (not "deg_free")

[erincperry]

oh okay perfect! I will try that. Thank you!

[nuripark10]

Hi Dr. McGowan, I have got my degrees of freedom as well as rmse, but do we need to do the "final_spec" just like we did in lab 4?

[LucyMcGowan]

Nope, you can skip that step!

[peytontarry]

@LucyMcGowan I'm also having trouble getting my code to run for this exercise. My code looks a little different from Erin's, but I think the error message that I'm receiving is similar. Do you have any pointers on how to fix mine? Do I need to create an object "grid" and add in a fit_resamples line like above?

[peytontarry]

@LucyMcGowan In case it's helpful, this is my recipe code!

[LucyMcGowan]

There’s a small typo (step_ns() should be for age not wage)

[peytontarry]

@LucyMcGowan Thank you so much!