Skip to content

Loading Smellscape Data

Loading example data

An example dataset is included with SmellscaPy to help you explore its features right away. Use the load_example_data() function to load the example dataset:

# Load example dataset
df = load_example_data()

The example dataset includes the following columns:

  • StudyID: Unique identifier of the study
  • ParticipantID: Unique participant ID
  • LocationID: Name of the city where the study was conducted
  • pleasant, present, light, engaging, unpleasant, absent, overpowering, detached: perceptual quality attributes

In addition, the dataset containes several optional variables:

  • How long have you been in your office without leaving? – Duration of continuous stay in the office
  • How many people are currently present in your office room? – Number of people sharing the same office
  • How would you rate your current mood? – Self-reported mood
  • In this moment, how productive do you feel? – Self-reported productivity
  • Smell source – Dominant category of perceived smell source

These data were collected in Bolzano, Italy, during an office-based experimental campaign in 2025 involving 17 participants.

# Display basic information about the example dataset
print(f"Dataset shape: {df}")
print(f"Columns ({len(df.columns)}): {list(df.columns)}")

Loading your data

If you have your own smellscape survey data, you can load it using pandas and then process it with SmellscaPy.

# Load data from a CSV file
my_data = pd.read_csv('path/to/your/data.csv')

To use smellscapy, your input dataset must contain the following mandatory columns:

  • ResearcherID
  • RecordID
  • LocationID
  • pleasant, present, light, engaging, unpleasant, absent, overpowering, detached

Make sure that the column names match exactly as listed above. If your dataset uses different names, you will need to rename the columns before using SmellscaPy.

Additional requirements:

  • No missing values are allowed in these columns.
  • The values for the eight perceptual attribute columns must be numeric, ranging from 1 (corresponding to "Strongly disagree") to 5 (corresponding to "Strongly agree).

Besides these required fields, you are free to include any other additional columns in your dataset as needed.

Please, refer to DataExample.csv for a reference dataset.