Working with Raw Biomechanical Data in R

Guide for analyzing biomechanical data directly from parquet files without the LocomotionData R package.

Loading Raw Data

library(arrow)
library(dplyr)
library(ggplot2)

# Load phase-indexed data
data <- read_parquet("converted_datasets/umich_2021_phase.parquet")

# Examine structure
glimpse(data)
str(data)

Data Structure

Raw parquet files contain: - subject: Subject identifier - task: Locomotion task (level_walking, incline_walking, etc.) - phase: Gait cycle percentage (0-149, representing 0-100%) - Biomechanical variables: knee_flexion_angle_ipsi_rad, hip_moment_contra_Nm, etc.

Basic Analysis

# Filter for specific subject and task
subject_data <- data %>%
  filter(subject == "SUB01", task == "level_walking")

# Calculate mean patterns
mean_patterns <- subject_data %>%
  group_by(phase) %>%
  summarise(across(where(is.numeric), mean, na.rm = TRUE))

# Plot knee angle
ggplot(mean_patterns, aes(x = phase, y = knee_flexion_angle_ipsi_rad)) +
  geom_line() +
  labs(title = "Mean Knee Flexion Angle",
       x = "Gait Cycle %", 
       y = "Angle (radians)")

Advanced Processing

# Calculate range of motion
rom_data <- subject_data %>%
  group_by(subject, task) %>%
  summarise(
    knee_rom = max(knee_flexion_angle_ipsi_rad) - min(knee_flexion_angle_ipsi_rad),
    .groups = "drop"
  )

# Group comparisons
group_comparison <- data %>%
  filter(task == "level_walking") %>%
  group_by(subject, phase) %>%
  summarise(mean_knee = mean(knee_flexion_angle_ipsi_rad, na.rm = TRUE), 
            .groups = "drop")

For more structured analysis, consider using the LocomotionData R package.