There is a super simple way to count all your results, in a row for example using value_count()
- "Return a Series containing counts of unique rows in the DataFrame."
Lets take this Series:
df['Hobbyist']
The above takes the Hobbyist
columne from the Stackoverflow dev survey. The output is below:
0 Yes
1 No
2 Yes
3 No
4 Yes
...
88878 Yes
88879 No
88880 No
88881 No
88882 Yes
Name: Hobbyist, Length: 88883, dtype: object
This is now a series and it has 88,882 entries, couting those could require a loop or a function, with pandas it's super simple.
df['Hobbyist'].value_counts()
Yes 71257
No 17626
Name: Hobbyist, dtype: int64
Easy pease!