"Group By" - lesson improvements

[mcdperera]

Hello dear maintainers,

I had the chance to teach Python session for beginners. While I was doing, I figure it out the existing “Group By” lesson is bit hard to understand.

So, I modified the section using same data set with different point of view. Here is my example of doing “Group By”.

*************************************** Start ***********************************

Group By: split-apply-combine

• Any groupby operation involves one of the following operations on the original object. They are −

  • Splitting the Object
  • Applying a function
  • Combining the results

• Split-apply-combine technique

Source: https://cmdlinetips.com/2018/02/introduction-to-split-apply-combine-with-pandas/

• Used "gapminder_all.csv" csv file.
  • Index using continent column.
  • Use Pandas "groupby()" method using continent column.

data = pd.read_csv('data/gapminder_all.csv', index_col=["continent"])
subset = data.groupby('continent')
print(subset)

• Above output have duplicate values.
• By using 'count()' method in Pandas we can get the exact record count (number of countries).

print(subset["country"].count())

• Use Pandas describe() method to do statistical analysis based on continents.

print(subset.groupby('continent').describe())

*************************************** End ***********************************

Conclusions

If you think the above mentioned suggestions could help, please do necessary changes to the lesson.

[eldobbins]

I also have had trouble teaching groupby - mainly because I didn't understand it well myself.

The problems I had with the example that calculates the wealth index:

• I can't see quickly the meaning of the second line. I had an easier time understanding groupby if I could predict what the answer would be
• The numbers output are a jumble (might be better sorted)
• Output is long (screen real-estate)

I like this new example because it solves those problems. I love the figure. However, I don't understand a couple of the outputs.

• print(subset) returns <pandas.core.groupby.generic.DataFrameGroupBy object at 0x120a86a30>. I don't mind that. It was helpful to me to see that I couldn't look at a groupby object directly. But is that what you meant?
• print(subset.groupby('continent').describe()) returned an AttributeError for me

'DataFrameGroupBy' object has no attribute 'groupby'

• the output of 'describe' overwhelms me. I would like to change the last line to a command that returns an easily understood DataFrame

subset.mean()

I will write a PR if I can get clarification about those things.

[vinisalazar]

I agree with the arguments presented by @eldobbins that the example that @mcdperera provided is better than the current one.

• print(subset) returns <pandas.core.groupby.generic.DataFrameGroupBy object at 0x120a86a30>. I don't mind that. It was helpful to me to see that I couldn't look at a groupby object directly. But is that what you meant?

I guess it would be nice to include a sentence or two about DataFrameGroupBy objects, and perhaps a link to the Pandas docs entry on these.

• print(subset.groupby('continent').describe()) returned an AttributeError for me

I'm guessing that OP meant subset.describe() instead? Seeing as the object is already grouped by continent.

• the output of 'describe' overwhelms me. I would like to change the last line to a command that returns an easily understood DataFrame

subset.mean()

I will write a PR if I can get clarification about those things.

I agree that subset.mean() would be more illustrative than .describe() in this example.

@eldobbins if you are still inclined to write a PR, I will be happy to review it or assist you (except for actually merging, that's up to @alee 😊 )

Best, V

[eldobbins]

I'd be obliged if you could do it. That was a long time ago for me! And you seem to understand what I was getting at.

Liz