Activity: How People Spend Their Day

Clean a real survey extract, then compare chores and leisure. About 45 minutes.

The Data Science Training Intensive’s spreadsheets track cleans an extract of the American Time Use Survey (ATUS) in Google Sheets. You have the same file, and now you have Python. Each row is one person’s diary of a single day, with the minutes they spent on chores, leisure and more. The file has mistakes in it, on purpose.

The question: how do people split their day between household chores and leisure, by number of children and by work status?

You’ll clean the file twice. First with plain Python, the tools from the intro session. Then with Polars, a library built for tables, which works on a whole column at once. Then you’ll answer the question and draw one chart.

NoteYour finish line

You’re done when you have all of these:

• 2,000 rows, and every Case ID is unique
• Sex and Education level cleaned
• Chores and Leisure columns
• The table of averages by number of children
• One chart
• atus_clean.csv saved

Before you start

You need the dss-python project and JupyterLab from the Setup page. Using Colab instead? Jump to the Colab box below.

1. Add Polars

You only need Polars from Part 2, so any time before then works. Open a terminal in your dss-python folder. JupyterLab has one: File → New → Terminal. Run:

uv add polars

If the notebook is already open, choose Kernel → Restart Kernel… so it can find Polars.

Using conda? Run this in your dss-python environment instead:

conda install -c conda-forge --override-channels polars

2. Download the learner notebook

Get it the same way you got the session notebook: Download the activity notebook. Your browser saves it in your Downloads folder. Drag it onto JupyterLab’s file list in the left sidebar, or click the Upload button (↑) above the list. Then double-click it.

3. The data comes with the notebook

The notebook’s first code cell downloads the two data files into the notebook’s folder, so there’s nothing to do. Want the files yourself? Save them in the same folder as the notebook: atus_working.csv, the file you’ll clean, and atus_raw.csv, the original survey codes.

TipOn Colab instead?

Open the learner notebook in Colab, sign in, and click Copy to Drive. Work in the tab titled Copy of activity-time-use.ipynb. Polars is already installed there, and the first code cell downloads the data.

What you’ll do

Work down the notebook from the top. Each section starts with a short explanation. Where a cell holds only a # comment, write the code under it, then run it with Shift-EnterShift-Enter.

Part 1: Plain Python

1. Read the file. The csv module turns each row into a dictionary.
2. Count the values in a column. Write count_values, and spot the values that shouldn’t be there.
3. A function that cleans a value. Write clean_sex, and fix every row.
4. Find an impossible age. One person has an age that can’t be right.

That was a lot of looping for one column. Part 2 does the same work, and more, a whole column at a time.

Part 2: Polars

5. Load the data with Polars. One line reads the whole file into a table.
6. Fix the text columns. Clean Sex and Education level for every row at once.
7. Codes, names and missing values. Give columns readable names, turn survey codes into labels, and mark impossible values as missing.
8. Duplicates. Remove repeated rows, then settle a Case ID that has two ages by checking the source file.
9. Chores and leisure. Add up the minutes for each person.
10. Answer the question. Average chores and leisure by number of children, then by labor force status.
11. One chart. Average chores by number of children.
12. Save your work. Write the clean data to atus_clean.csv.

Check your numbers

Same numbers as the spreadsheets track? Compare them at the wrap-up.

Rows and cleaning

After	Rows
Reading the file (section 5)	2,006
unique(), which removes exact copies (section 8)	2,001
Dropping the row with the wrong age (section 8)	2,000, and every Case ID is unique

• Sex has two values: Female and Male.
• Education level has 8 values: No diploma, High School, Some college-no degree, Associate’s, Bachelor’s, Master’s, Professional and Doctoral.
• Missing values at df.null_count() in section 7: Age 1, Sleep (mins) 3, Work hours per week 839, Metropolitan 233.

Chores and leisure

Averages over everyone: chores 74.1 minutes and leisure 252.2 minutes a day.

By number of children (3 means 3 or more):

Children	People	Chores (mins)	Leisure (mins)
0	1137	65.7	297.9
1	323	66.2	206.9
2	360	85.9	187.7
3 or more	180	117.4	173.5

By labor force status, biggest group first:

Status	People	Chores (mins)	Leisure (mins)
Employed, at work	1181	62.9	200.9
Not in labor force	684	91.5	337.1
Unemployed, looking	72	87.4	273.3
Employed, absent	55	71.1	258.6
Unemployed, on layoff	8	140.4	326.0

About the data

Each row is one respondent’s diary for a single day, 24 hours. The activity columns end in (mins), and hold the minutes spent on each activity that day.

Column	Meaning	Values and codes
Case	Respondent ID (TUCASEID in the survey)	A long number, one per person
Diary Year	Year of the diary day	2003–2023
Diary Day	Day of the week	Monday to Sunday
Metropolitan	Lives in a metropolitan area	1 Metropolitan, 2 Non-metropolitan, 3 Not identified, -1 Blank
Education level	Highest education	Text, like Bachelor’s
Age	Age in years	15–85. 80 and 85 are top codes: older ages are grouped there.
Sex	Sex of the respondent	Female or Male
TRCHILDNUM	Children under 18 in the household	0–8
Labor force status	Work status	Employed, at work. Employed, absent. Unemployed, looking. Unemployed, on layoff. Not in labor force.
Enrolled in school	In school or not	Yes, No, or Blank (not asked)
TEHRUSLT	Usual weekly work hours	Hours. -1 means blank (not asked), -4 means hours vary.
Sleep (mins)	Minutes asleep that day	The text NA means missing.
Interior cleaning, Laundry, Food and drink prep, Kitchen clean-up	Minutes on each chore	These four add up to Chores.
Reading, Playing games, TV and Movies, Socializing	Minutes on each leisure activity	These four add up to Leisure.
Eating and Drinking, Grocery Shopping	Minutes eating, and shopping for food	Counted as neither chores nor leisure.

atus_raw.csv has the same 2,000 people with the survey’s own variable names and numeric codes, like TUCASEID, TESEX and TELFS. It’s the source of truth when two rows disagree.

Source: American Time Use Survey (U.S. Bureau of Labor Statistics, bls.gov/tus), via ICPSR study 36268. The DSTI team sampled 2,000 respondents from diary years 2003–2023.

Results describe this sample, not the US. The file has no survey weights, so the averages don’t estimate what Americans do overall.

Stuck?

• The filled notebook has every answer: download the filled notebook, or open the filled notebook in Colab.
• The cheat sheet has JupyterLab keys, syntax and common errors on one page.
• Help with setup or access: go.ncsu.edu/getdatahelp.

What you see	What to do
ModuleNotFoundError for polars	Run uv add polars in dss-python, then choose Kernel → Restart Kernel… and run the cell again.
FileNotFoundError for atus_working.csv	Run the Get the data cell near the top first. Or save the CSV files in the notebook’s folder, as in step 3 of Before you start.

Numbers checked with Polars 2.0.0.