Python Coding Challenge

Expected Time: 30 minutes

Thanks so much for your interest in the Social Interaction & Neural Computation Laboratory! Below, you will find a coding challenge that will mimic a few simple everyday problems we encounter in the lab. This challenge is designed to assess your ability to solve problems and analyze data using Python and relevant libraries (e.g., pandas, numpy, matplotlib). We will evaluate the code based your coding style, efficiency, and creativity in solving the problems — it is okay if you do not know how to solve all of the problems. We will be looking at your thought process and approach to the problems. Please comment your code to explain your thought process. I also encourage using external resources (e.g., Google, StackOverflow, ChatGPT) to help you, but please indicate where and how you used these resources in your code.

You are provided with a comma-delimited file (dyad_neural_data.csv) and a Python script (challenge.py). Your task is to complete the functions within the script by following the prompts below (do not edit anything after if __name__ == "__main__": as these will help make sure you are on track). You may use pandas, numpy, and matplotlib (or seaborn) to solve the problems.

Once completed, please email Shawn Rhoads your final script as LASTNAME_FIRSTNAME_challenge.py (with your name inserted in the filename). Before sending, please verify that running the following command works for you:

python LASTNAME_FIRSTNAME_challenge.py

Please download the challenge materials here:

• Data: dyad_neural_data.csv
• Script: challenge.py

Dataset Description

The dataset (dyad_neural_data.csv) contains mock neural recordings for subjects engaged in dyadic interactions, recorded every second for 60 seconds. Some subjects are matched with more than one partner. See example below.

• subj_id: Unique subject identifier
• dyad_id: Identifier for the dyadic interaction
• time: Time in seconds (0-60)
• power: Power value of neural activity (relative to a baseline)

🚨 Hint: Subject 5 is in two dyads: Dyad C (with Subject 6) and Dyad D (with Subject 7).

Part 1: Data Organization

1. Identify Subjects with Multiple Dyad Matches

• Write a function find_multiple_partners(df) that returns a list of subjects (subj_id) who appear in more than one dyad_id.

2. Filter Specific Time Range

• We will often focus only on epochs of the data where there is an event of interest. Filter the data to only include timepoints 24-45. Write a function select_timepoints(df, xmin, xmax) that returns a new dataframe that only includes rows where time is between 24 seconds (xmin) and 45 seconds (xmax) for each subject while retaining all other columns.

Part 2: Data Manipulation

3. Compute Mean Power

• Write a function compute_mean_power(df) that filters the data to only include timepoints 30-70 (using your function from earlier) and returns a new dataframe with the mean power for each subj_id for this time range.

4. Standardize Power Values

• Write a function zscore_power(df) that standardizes the power column over time for each subject by subtracting the mean and dividing by the standard deviation. The output should return a new dataframe with an added column called zpower.

5. Apply a Custom Function to Scale Data

• Write a function scale_power(df) that applies a lambda function that scales power values by the maximum value in the sample. The output should return a new dataframe with an added column called scaled_power.

Part 3: Data Analysis and Visualization

6. Detect Dyads with High Variability in Power

• Write a function detect_high_variability_dyads(df, sigma=.25) to identify dyads where the standard deviation of power across subjects exceeds a threshold (sigma=0.25) and returns a list of those subjects.

7. Extract and Plot Activity as a Function of Time

• Write function plot_power_x_time(df) to plot the z-scored timeseries for all subjects, color each dyad uniquely and add a legend.
• ✨ Bonus: only plot the timepoints 33-68 using your function from earlier.

8. Compute and Plot Inter-Subject Correlation

• Write a function plot_correlation_heatmap(df) to compute the Pearson r correlation of power between subjects’ timeseries (exclude dyad “D”) and plot a heatmap of the subject × subject correlation matrix where the colormap corresponds to the correlation coefficient. Also, mask out the upper triangle.
• 🚨 Hint: The correlation matrix should be a 14 x 14 matrix.
• ✨ Bonus: Draw a border around actual partner pairs (e.g., subject 1 and subject 6 are not partners, but subject 1 and subject 2 are). Annotate these heatmap cells with their respective correlation coefficients.