microt-compliance

Introduction

This repository include Python scripts that transform raw sensor data collected from Android mobile devices into features of interest for analysis.

Current research topic: identify contextual prompt-level factors that predict response or no response in microEMA.

Statistical methodology: multi-level modeling

Outcomes: compliance rate

Predictors:

• Within-person or prompt level:

  • time of the day
  • day of the week
  • day in study
  • activity level
  • battery level
  • location

• Between-person or person level

  • age
  • gender
  • study mode

Exploratory Discussions on Factors

Detailed discussions on factors can be found here.

Features Overview

Smartphone

Outcome	Variable Type	Data Source
Answer Status	Categorical (Completed, CompletedThenDismissed, PartiallyCompleted, Started/NeverStarted, NeverPrompted, OverwrittenByDaily)	./logs/PromptResponses.log.csv

Feature	Level	Effect Type	Variable Type	Data Source
Day of the Week	Level 1 (Within-person or prompt level)	Random	Categorical (Mon-Sat: 0-6)	./logs/.../PromptResponses.log.csv
Time of the Day	Level 1 (Within-person or prompt level)	Random	Categorical (morning, afternoon, evening/night)	./logs/.../PromptResponses.log.csv
Days in Study	Level 1 (Within-person or prompt level)	Random	Numeric (numeric value of day from the first day)	./logs (start from the first date of created folder)
Battery Level	Level 1 (Within-person or prompt level)	Random	Numeric (Battery%)	./data/.../Battery.##.event.csv
Charging Status	Level 1 (Within-person or prompt level)	Random	Binary (True/False)	./data/.../Battery.##.event.csv
Location (LOC)	Level 1 (Within-person or prompt level)	Random	[Latitude, Longitude]	./data/.../GPS.csv
Phone Lock	Level 1 (Within-person or prompt level)	Random	Binary (Phone Locked/Phone Unlocked)	./data/.../AppEventCounts.csv
Last Phone Usage Duration	Level 1 (Within-person or prompt level)	Random	Numeric (minutes)	./data/.../AppEventCounts.csv
Screen Status	Level 1 (Within-person or prompt level)	Random	Binary (Screen On/Screen Off)	./logs/.../SystemBroadcastReceiver.csv
Wake/Sleep Time	Level 1 (Within-person or prompt level)	Random	Local time (2021-01-01 06:30:00 CST)	daily report

Smartwatch

Outcome	Level	Effect Type	Variable Type	Data Source
Compliance Rate	Level 1 (Within-person or prompt level)	Random	Numeric	./logs-watch/PromptResponses.log.csv
	Level 2 (Between-person or person level)	Random

Feature	Level	Effect Type	Variable Type	Data Source
Day of the Week	Level 1 (Within-person or prompt level)	Random	Categorical (Mon-Sat: 0-6)	./logs-watch/.../PromptResponses.log.csv
Time of the Day	Level 1 (Within-person or prompt level)	Random	Categorical (morning, afternoon, evening/night)	./logs-watch/.../PromptResponses.log.csv
Days in Study	Level 1 (Within-person or prompt level)	Random	Numeric (numeric value of day from the first day)	./logs-watch (start from the first date of created folder)
Battery Level	Level 1 (Within-person or prompt level)	Random	Numeric (Battery%)	./data-watch/.../Battery.##.event.csv
Location (LOC)	Level 1 (Within-person or prompt level)	Random
Activity Level (ACT)	Level 1 (Within-person or prompt level)	Random

Code Usage for Feature Matrix Generation

• This project has been wrapped up as a Pypi package. Use pip to install.
• Clone this project and run locally.

#!python
python main_ema.py [intermediate_participant_path] [output_dir_path] [date_in_study] [decryption_password]

e.g., python main_ema.py G:...\intermediate_file\participant_id C:...\output_folder 2021-01-01 password

Special Notice

• Delete misc folder before running code, if new participants' intermediate folder has been created.