Improve embedding of event data derived from external functions

[vincentvanhees]

Currently GGIR facilitates the embedding of external functions, with the step detection algorithm as a clear example. However, the same would be true for algorithms that facilitate the detection of other event type data such as falls, tremors, scratch, seizures, .....

However, at the moment the output is treated as just another signal metric similar to acceleration metrics. Only a little bit of effort was made to tailor the reporting to the unique nature of these events.

In this project I will expand GGIR's ability to utilize and describe event data beyond treating it as just another signal metric.

Focus points include:

• Alternative imputation approach.
• Summing rather than averaging per day, per qwindow, and per acceleration level.
• Event bout detection.

UPDATE (7 August 2024):

• Work has been completed a year ago, but I am still waiting for client to publish their research after which this can be merged.

UPDATE (20 November 2024):

• PR generated to merge first half of work with new output suppressed
• Outstanding tasks for 2025:
  • Add dedicated vignette to summarise the new functionality
  • Unsuppress new output
  • Address redundancy in MXLX / M5L5 calculations as discussed in redundant M5L5 code #312.
  • Review parameter and output variable names and implement simplifications where possible.