PMA Analytics for Python

The repository is named pma2020-analytics2 because this is the second pass at writing analytics tools. The first pass was done in R. The Python package is simply called analytics. The functionality is provided as a command-line tool.

Note: the repository name and the package name are different!

Some of the data that is extracted:

• Specific XML tags from the submitted instance
• Total active screen time during the whole survey
• Short break time during the whole survey
• File sizes of photos, submission, and log
• Total swiping events

Also, five data points for each known prompt in the log are recorded. Each column name suffix and description is below.

• _c for the number of times a constraint/required was invoked on the prompt.
• _t for the active screen time spent on the prompt.
• _v for the total number of visits to the prompt.
• _d for the total number of times the answer changed on the prompt.
• _b for short break time associated with the prompt.

All times are in millseconds.

Pre-requisites

PMA Analytics makes use of Python 3.6. Install Python 3.6.

Installation

Install via pip with

python3 -m pip install https://github.com/jkpr/pma2020-analytics2/zipball/master

Examples

All required arguments are named to follow the same pattern as ODK Briefcase.

Example usage:

python3 -m analytics.condense --storage_directory ~/Documents/odkbriefcase/ --form_id HQ-rjr1-v25 --export_directory . --export_filename hq-out.csv

A JSON file can be supplied through the --lookup option of the condense command-line interface. The file should have the proper format: a list of JSON objects with properties

• form_id (string) form id
• form_title (string) form title
• prompts (list of string) prompts in the log.txt files
• tags (list of string) names of XML tags from submission.xml files

Put something in prompts in order to get information about number of visits and time spent at that prompt. Put something in tags to get the actual value in the submission, e.g. <your_name>Jane Doe</your_name>

An example is given below.

[
  {
    "form_id": "HQ-rjr1-v12",
    "form_title": "RJR1-Household-Questionnaire-v12",
    "prompts": [
      ...
      "hh_duplicate_check",
      "duplicate_warning",
      "resubmit_reasons",
      "duplicate_warning_hhmember",
      "available",
      "consent_start",
      "consent",
      "begin_interview",
      ...
    ],
    "tags": [
      ...
      "your_name",
      "start",
      "end",
      "deviceid",
      "HHQ_result"
      ...
    ]
  }
]

Supported ODK forms

The analytics.formdata subpackage has a command line interface to display the supported form titles and form ids. Use:

python3 -m analytics.formdata

Then pipe to grep to filter results. For example, in order to see the supported forms for Uganda round 5, use

python3 -m analytics.formdata | grep UGR5

Understanding the logging

As analytics runs, it emits logging messages. The standard levels, in order of increasing severity, are DEBUG, INFO, WARNING, and ERROR. Analytics uses all of these to convey specific meaning.

DEBUG

Information that is useful for debugging. Analytics uses DEBUG to say which instance (folder) is currently being analyzed.

INFO

High-level information about the running program.

• What form id is being analyzed.
• When the analysis completes

WARNING

When WARNING is about the logs, it typically means something with the potential to be problematic has occurred. Usually, however, this is not a cause for concern.

ERROR

This is used when something happens that prevents the analytics from completing its task.

• The analytics file column headers do not match what data is to be appended.
• Some problem with the file prevents its analysis (e.g. corrupt file).

Updates

python3 -m pip install https://github.com/jkpr/pma2020-analytics2/zipball/master --upgrade

Bugs

Submit bug reports to James Pringle at jpringleBEAR@jhu.edu minus the BEAR.