aggregating model summaries from list of models

Question

I am running same models several times on different subset of data, for example,

  data("mtcars")
  head("mtcars")

  table(mtcars$cyl, useNA = "ifany")

I am fitting the model (mpg ~ hp + wt) for each cylinder type.

 foo <- mtcars %>%
          group_by(cyl) %>%
             nest() %>%
               mutate(model= map(data, rms::ols(mpg ~ hp + wt, data = .)))

When I try this

  foo %>%
    {map(.$model, summary)}

I get the model summary from all three models but this does not include the estimates for intercept.

I am not interested in fitting the model using lm function, I am specifically interested in solutions involving fitting rms::ols() function.

Any help on aggregating the summaries from multiple models using rms::ols() and that includes estimates for intercept, will be very helpful. Thanks.

Answer 1

As stated in the comments, your question is more about having a custom summary function.

Note that what you called "induvidual model summary, foo$model" is not a summary (as in output for a summary function), is just the default printing method of a rms object.

You can use summary.lm():

foo %>%
  {map(.$model, summary.lm)}

Extra: if you don't want your models saved as a tibble, you can use group_split -> map, instead of group_by -> nest -> map:

foo <- mtcars %>%
  group_split(cyl) %>%
  map(~ rms::ols(mpg ~ hp + wt, data = .x))

Now, foo is a list and you can do more naturally:

map(foo, summary.lm)