Ignoring NA_real_ when calculating with group_by

Question

I need to do some calculations regarding rows with a same ID (mean, max and sum)

For instance, from all individuals in a same family, which one has the greatest education level:

yst = Years of study of this individual

idf = Family identification number for this individual

maxyst = Years of study of this most educated person in this individual's family

I tried to use mutate() and group_by(), but when there is one NA_real_ row the whole group gets NA. What I want to do is to ignore any idf in which ALL rows are is.na(yst). But if I have at least one !is.na(yst), I wish to ignore all the NA and keep the calculation within the valid rows in that idf.

This is my current code:

df %<>% 
  mutate(df, yst = anosest) %>%
  mutate(df, yst = case_when(
    is.na(anosest) ~ 0,
    TRUE ~ yst)) %>%
  group_by(idf) %>%
  mutate(maxyst = max(yst)) %>%
  ungroup()

Tried a lot of different approaches using a combination of logical values, but I couldn't figure it out.