groups attribute of a groupby object question

[smw10c]

I am having trouble understanding the following code. Note, auto is a DataFrame that contains the variables yr (year) and mpg (miles per gallon):

splitting=auto.groupby(‘yr’)
type(splitting.groups)
for group_name, group in splitting:
    avg=group[‘mpg’].mean()
    print(group_name,avg)

Here is what I understand: we are saving a groupby object to "splitting" that is grouped by year. Next, we see that the type of splitting.groups is a dictionary. We iterate over the key value pairs in splitting, obtain an average, and print the key along with it's average mpg.

What is confusing me is the line "avg=group['mpg'].mean()." From what I've learned about indexing, I've never seen a value being able to be used at the beginning of the index. Since group is the value in the key value pairs, how does Python know that group['mpg'] refers to the mpg column?

[zivoni]

Your group_name and group is not a key: value pair from grouping.groups. Iterating over dataframe's groupby yields unique values of grouping variable and appropriate subset of the original dataframe. In your case group_name is yr value and group is a subset of the original auto dataframe, basically auto[auto.yr==group_name]. So group['mpg'] is column with mpg values (again, its basically part of auto['mpg'] column).

You could get similar result without groupby with:

for yr in auto.yr.unique():  # yr is like your group_name
    group = auto[auto.yr==yr]
    avg = group['mpg'].mean()
    print(yr, avg)

But usually you would use just auto.groupby('yr')['mpg'].mean() without any explicit looping.

[smw10c]

Your group_name and group is not a key: value pair from grouping.groups. Iterating over dataframe's groupby yields unique values of grouping variable and appropriate subset of the original dataframe. In your case group_name is yr value and group is a subset of the original auto dataframe, basically auto[auto.yr==group_name]. So group['mpg'] is column with mpg values (again, its basically part of auto['mpg'] column).

That makes a lot of sense. Took me a few times to read over to understand but I got it all now lol. Thanks for the thorough explanation.