Using Pandas

Last updated on 2024-05-17 | Edit this page

Estimated time: 30 minutes

Overview

Questions

  • How can I work with subsets of data in a pandas DataFrame?
  • How can I run summary statistics and sort columns of a DataFrame?
  • How can I save DataFrames to other file formats?

Objectives

  • Select specific columns and rows from pandas DataFrames.
  • Use pandas methods to calculate sums and means, and to display unique items.
  • Sort DataFrame columns (pandas series).
  • Save a DataFrame as a CSV or pickle file.

Pinpoint specific rows and columns in a DataFrame

If you don’t already have all of the CSV files loaded into a DataFrame, let’s do that now:

PYTHON 

import glob
import pandas as pd

dfs = [] 

for csv in sorted(glob.glob('data/*.csv')):
    year = csv[5:9] 
    data = pd.read_csv(csv) 
    data['year'] = year 
    dfs.append(data)

df = pd.concat(dfs, ignore_index=True)

df.head(3)
branch address city zip code january february march april may june july august september october november december ytd year
0 Albany Park 5150 N. Kimball Ave. Chicago 60625.0 8427 7023 9702 9344 8865 11650 11778 11306 10466 10997 10567 9934 120059 2011
1 Altgeld 13281 S. Corliss Ave. Chicago 60827.0 1258 708 854 804 816 870 713 480 702 927 787 692 9611 2011
2 Archer Heights 5055 S. Archer Ave. Chicago 60632.0 8104 6899 9329 9124 7472 8314 8116 9177 9033 9709 8809 7865 101951 2011

Use tail() to look at the end of the DataFrame

We’ve seen how to look at the first rows in your DataFrame using .head(). You can use .tail() to look at the final rows.

PYTHON 

df.tail(3)
branch address city zip code january february march april may june july august september october november december ytd year
960 Brighton Park 4314 S. Archer Ave. Chicago 60632.0 1394 1321 1327 1705 1609 1578 1609 1512 1425 1603 1579 1278 17940 2022
961 South Chicago 9055 S. Houston Ave. Chicago 60617.0 496 528 739 775 587 804 720 883 681 697 799 615 8324 2022
962 Chicago Bee 3647 S. State St. Chicago 60609.0 799 543 709 803 707 931 778 770 714 835 718 788 9095 2022

Slicing a DataFrame

We can use the same slicing syntax that we used for strings and lists to look at a specific range of rows in a DataFrame.

PYTHON 

df[50:60] #look at rows 50 to 59
branch address city zip code january february march april may june july august september october november december ytd year
50 Near North 310 W. Division St. Chicago 60610.0 11032 10021 12911 12621 12437 13988 13955 14729 13989 13355 13006 12194 154238 2011
51 North Austin 5724 W. North Ave. Chicago 60639.0 2481 2045 2674 2832 2202 2694 3302 3225 3160 3074 2796 2272 32757 2011
52 North Pulaski 4300 W. North Ave. Chicago 60639.0 3848 3176 4111 5066 3885 5105 5916 5512 5349 6386 5952 5372 59678 2011
53 Northtown 6435 N. California Ave. Chicago 60645.0 10191 8314 11569 11577 10902 14202 15310 14152 11623 12266 12673 12227 145006 2011
54 Oriole Park 7454 W. Balmoral Ave. Chicago 60656.0 11999 11206 13675 12755 10364 12781 12219 12066 10856 11324 10503 9878 139626 2011
55 Portage-Cragin 5108 W. Belmont Ave. Chicago 60641.0 9185 7634 9760 10163 7995 9735 10617 11203 10188 11418 10718 9517 118133 2011
56 Pullman 11001 S. Indiana Ave. Chicago 60628.0 1916 1206 1975 2176 2019 2347 2092 2426 2476 2611 2530 2033 25807 2011
57 Roden 6083 N. Northwest Highway Chicago 60631.0 6336 5830 7513 6978 6180 8519 8985 7592 6628 7113 6999 6082 84755 2011
58 Rogers Park 6907 N. Clark St. Chicago 60626.0 10537 9683 13812 13745 13368 18314 20367 19773 18419 18972 17255 16597 190842 2011
59 Roosevelt 1101 W. Taylor St. Chicago 60607.0 6357 6171 8228 7683 7257 8545 8134 8289 7696 7598 7019 6665 89642 2011

Look at specific columns

To work specifically with one column of a DataFrame we can use a similar syntax, but refer to the name the column of interest.

PYTHON 

df['year'] #look at the year column

OUTPUT 

0      2011
1      2011
2      2011
3      2011
4      2011
       ... 
958    2022
959    2022
960    2022
961    2022
962    2022
Name: year, Length: 963, dtype: object

We can add a second square bracket after a column name to refer to specific row indices, either on their own, or using slices to look at ranges.

PYTHON 

print(f"first row: {df['year'][0]}") #use double quotes around your fstring if it contains single quotes
print('rows 100 to 102:') #add a new print statement to create a new line
print(df['year'][100:103])

OUTPUT 

first row: 2011
rows 100 to 102:
100    2012
101    2012
102    2012
Name: year, dtype: object

Columns display differently in our notebook since a column is a different type of object than a full DataFrame.

PYTHON 

type(df['year'])

OUTPUT 

pandas.core.series.Series

Summary statistics on columns

A pandas Series is a one-dimensional array, like a column in a spreadsheet, while a pandas DataFrame is a two-dimensional tabular data structure with labeled axes, similar to a spreadsheet. One of the advantages of pandas is that we can use built-in functions like max(), min(), mean(), and sum() to provide summary statistics across Series such as columns. Since it can be difficult to get a sense of the range of data in a large DataFrame by looking over the whole thing manually, these functions can help us understand our dataset quickly and ask specific questions.

If we wanted to know the range of years covered in this data, for example, we can look at the maximum and minimum values in the year column.

PYTHON 

print(f"max year: {df['year'].max()}")
print(f"min year: {df['year'].min()}")

OUTPUT 

max year: 2022
min year: 2011

Summarize columns that hold string objects

We might also want to quickly understand the range of values in columns that contain strings, the branch column, for example. We can look at a range of values, but it’s hard to tell how many different branches are present in the dataset this way.

PYTHON 

df['branch']

OUTPUT 

0         Albany Park
1             Altgeld
2      Archer Heights
3              Austin
4       Austin-Irving
            ...      
958         Chinatown
959          Brainerd
960     Brighton Park
961     South Chicago
962       Chicago Bee
Name: branch, Length: 963, dtype: object

We can use the .unique() function to output an array (like a list) of all of the unique values in the branch column, and the .nunique() function to tell us how many unique values are present.

PYTHON 

print(f"Number of unique branches: {df['branch'].nunique()}")
print(df['branch'].unique())

OUTPUT 

Number of unique branches: 82
['Albany Park' 'Altgeld' 'Archer Heights' 'Austin' 'Austin-Irving'
 'Avalon' 'Back of the Yards' 'Beverly' 'Bezazian' 'Blackstone' 'Brainerd'
 'Brighton Park' 'Bucktown-Wicker Park' 'Budlong Woods' 'Canaryville'
 'Chicago Bee' 'Chicago Lawn' 'Chinatown' 'Clearing' 'Coleman'
 'Daley, Richard J. - Bridgeport' 'Daley, Richard M. - W Humboldt'
 'Douglass' 'Dunning' 'Edgebrook' 'Edgewater' 'Gage Park'
 'Galewood-Mont Clare' 'Garfield Ridge' 'Greater Grand Crossing' 'Hall'
 'Harold Washington Library Center' 'Hegewisch' 'Humboldt Park'
 'Independence' 'Jefferson Park' 'Jeffery Manor' 'Kelly' 'King'
 'Legler Regional' 'Lincoln Belmont' 'Lincoln Park' 'Little Village'
 'Logan Square' 'Lozano' 'Manning' 'Mayfair' 'McKinley Park' 'Merlo'
 'Mount Greenwood' 'Near North' 'North Austin' 'North Pulaski' 'Northtown'
 'Oriole Park' 'Portage-Cragin' 'Pullman' 'Roden' 'Rogers Park'
 'Roosevelt' 'Scottsdale' 'Sherman Park' 'South Chicago' 'South Shore'
 'Sulzer Regional' 'Thurgood Marshall' 'Toman' 'Uptown' 'Vodak-East Side'
 'Walker' 'Water Works' 'West Belmont' 'West Chicago Avenue'
 'West Englewood' 'West Lawn' 'West Pullman' 'West Town'
 'Whitney M. Young, Jr.' 'Woodson Regional' 'Wrightwood-Ashburn'
 'Little Italy' 'West Loop']

Use .groupby() to analyze subsets of data

A reasonable question to ask of the library usage data might be to see which branch library has seen the most checkouts over this ten + year period. We can use .groupby() to create subsets of data based on the values in specific columns. For example, let’s group our data by branch name, and then look at the ytd column to see which branch has the highest usage. .groupby() takes a column name as its argument and then for each group we can sum the ytd columns using .sum().

PYTHON 

df.groupby('branch')['ytd'].sum()

OUTPUT 

branch
Albany Park              1024714
Altgeld                    68358
Archer Heights            803014
Austin                    200107
Austin-Irving            1359700
                          ...   
West Pullman              295327
West Town                 922876
Whitney M. Young, Jr.     259680
Woodson Regional          823793
Wrightwood-Ashburn        302285
Name: ytd, Length: 82, dtype: int64

Sort pandas series using .sort_values()

The output for code above is another pandas series object. Let’s save the output to a new variable so we can then apply the .sort_values() method which allows us to view the branches with the most usage. The ascending parameter for .sort_values() takes True or False. We want to pass False so that we sort from the highest values down…

PYTHON 

circ_by_branch = df.groupby('branch')['ytd'].sum()
circ_by_branch.sort_values(ascending=False).head(10)

OUTPUT 

branch
Harold Washington Library Center    7498041
Sulzer Regional                     5089225
Lincoln Belmont                     1850964
Edgewater                           1668693
Logan Square                        1539816
Rogers Park                         1515964
Bucktown-Wicker Park                1456669
Lincoln Park                        1441173
Austin-Irving                       1359700
Bezazian                            1357922
Name: ytd, dtype: int64

Now we have a list of the branches with the highest number of uses across the whole dataset.

We can pass multiple columns to groupby() to subset the data even further and breakdown the highest usage per year and branch. To do that, we need to pass the column names as a list. We can also chain together many methods into a single line of code.

PYTHON 

circ_by_year_branch = df.groupby(['year', 'branch'])['ytd'].sum().sort_values(ascending=False)
circ_by_year_branch.head(5)

OUTPUT 

year  branch                          
2011  Harold Washington Library Center    966720
2012  Harold Washington Library Center    937649
2013  Harold Washington Library Center    821749
2014  Harold Washington Library Center    755189
2015  Harold Washington Library Center    694528
Name: ytd, dtype: int64

Use .iloc[] and .loc[] to select DataFrame locations.

You can point to specific locations in a DataFrame using two-dimensional numerical indexes with .iloc[].

PYTHON 

# print values in the 1st and 2nd to last columns in the first row
# '\n' prints a linebreak
print(f"Branch: {df.iloc[0,0]} \nYTD circ: {df.iloc[0,-2]}")

OUTPUT 

Branch: Albany Park 
YTD circ: 120059

.loc[] uses the same structure but takes row (index) and column names instead of numerical indexes. Since our df rows don’t have index names we would still use the default numerical index.

PYTHON 

# print the same values as above, using the column names
print(f"Branch: {df.loc[0,'branch']} \nYTD circ: {df.loc[0, 'ytd']}")

OUTPUT 

Branch: Albany Park 
YTD circ: 120059

Save DataFrames

You might want to export the series of usage by year and branch that we just created so that you can share it with colleagues. Pandas includes a variety of methods that begin with .to_... that allow us to convert and export data in different ways. First, let’s save our series as a DataFrame so we can view the output in a better format in our Jupyter notebook.

PYTHON 

circ_df = circ_by_year_branch.to_frame()
circ_df.head(5)
ytd
year branch
2011 Harold Washington Library Center 966720
2012 Harold Washington Library Center 937649
2013 Harold Washington Library Center 821749
2014 Harold Washington Library Center 755189
2015 Harold Washington Library Center 694528

Save to CSV

Next, let’s export the new DataFrame to a CSV file so we can share it with colleagues who love spreadsheets. The .to_csv() method expects a string that will be the name of the file as a parameter. Make sure to add the .csv filetype to your file name.

PYTHON 

circ_df.to_csv('high_usage.csv')

You should now see, in the JupyterLab file explorer to the left, the new CSV file. If you don’t see it, you can hit the refresh icon (it looks like a spinning arrow) above the files pane. You can double-click on the CSV to preview the full spreadsheet in a new Jupyter tab.

Save pickle files

Working with your data in CSVs (especially via tools like Microsoft Excel) can introduce reproducibility issues. For example, you’ll sometimes encounter character encoding problems, where certain characters in your dataset will no longer display properly after editing them in a spreadsheet software like Excel, and re-importing them to a pandas DataFrame.

One way to avoid issues like this is to save Python objects as pickles. Technically speaking, the Python pickle module serializes and de-serializes a Python object’s structure. In practical terms, pickling allows you to store Python objects (like DataFrames, lists, etc.) efficiently and without losing or corrupting your data.

You can save a DataFrame to pickle by using the to_pickle() method and using the filetype of pkl.

PYTHON 

circ_df.to_pickle('high_usage.pkl')

You can only “see” the data in a pickle file by reloading it into Python. This is a great way to save a DataFrame that you created in one JupyterLab session so that you can reload it later on, or share it with a colleague who’s familiar with Python.

PYTHON 

new_df = pd.read_pickle('high_usage.pkl')
new_df.head()

Finally, let’s save our full concatenated DataFrame to a pickle file that we can use later on in the lesson. We’ll save it in the data/ directory alongside our other data files.

PYTHON 

df.to_pickle('data/all_years.pkl')

Displaying rows and columns

How would you use slicing and column names to select the following subsets of rows and columns from the circulation DataFrame?

  1. The city column.
  2. Rows 10 to 20.
  3. Rows 20 to 30 from the zip code column.

PYTHON 

#1
df['city']

#2
df[10:21]

#3 
df['zip code'][20:31]

Using loc()

How would you use loc() to select rows 20 to 30 from the zip code column (the same rows as the last example in the challenge above)?

Tip: slices use “non-inclusive” indexing – so require you to ask for df[10:21] to see row 20, but loc() uses inclusive indexing.

PYTHON 

df.loc[20:30, 'zip code']

Unique items

How would you display:

  1. all of the unique zip codes in the dataset?
  2. the number of unique zip codes in the dataset?

PYTHON 


#1
df['zip code'].unique()

#2
df['zip code'].nunique()

Summary statistics and groupby()

We can apply mean() to pandas series’ in the same way we used sum(), min(), and max() above. How would you display the following?

  1. the mean number of ytd checkouts grouped by zip code?
  2. the mean number of ytd checkouts grouped by zip code, and sorted from smallest to largest?

PYTHON 

#1
df.groupby('zip code')['ytd'].mean()

#2
df.groupby('zip code')['ytd'].mean().sort_values()

Key Points

  • Use builtin methods .sum(), .mean(), unique(), and nunique() to explore summary statistics on the rows and colums in your DataFrame.
  • Use .groupby() to work with subsets of your dataset.
  • Sort pandas series with .sort_values().
  • Use .loc() and .iloc() to pinpoint specific locations in Pandas DataFrames.
  • Save DataFrames to CSV and pickle files using .to_csv() and .to_pickle().