CODEWITHSUNDEEP
rtidyversepurrr
Geet
Geet

Reputation: 2575

Purrr: Which nested dataframe encountered the error?

Here is a nested data. 

df1 <- tibble::tribble(
   ~A,        ~B,  ~group,
   4L,         1L,  "A",
   7L,         4L,  "A",
  NA_integer_, 1L,  "B",
  NA_integer_, 10L, "B")

df2 <- df1 %>% group_by(group) %>% nest()

I need to run lm using purrr:map. 

map(df2$data, ~lm(A~B, data=.x))

What's the best way to find out which nested data encountered an error i.e. how can I know that group B has an issue.

Can you suggest a solution may be using purrr's possibly or safely?

Upvotes: 2

Views: 147

Answers (1)

aosmith
aosmith

Reputation: 36086

You can use possibly() here. In my example I have it return NA if the model has an error.

First I make posslm, using otherwise to tell it what to return if there is an error after using lm.

posslm = possibly(lm, otherwise = NA)

Then you can make a new column of models with map in mutate. Once that's done, filter to NA rows for the new variable and then pull out the group.

mutate(df2, mod = map(data, ~posslm(A~B, data=.x))) %>%
    filter( is.na(mod) ) %>%
    pull(group)

[1] "B"

Upvotes: 5

Related Questions