From 48626811953b0b7a23646aae9432783378e461f1 Mon Sep 17 00:00:00 2001 From: Vincent van Hees Date: Thu, 7 Nov 2024 10:54:02 +0100 Subject: [PATCH] minor expansions to chapter 0 and 1 related to independent initiatives --- vignettes/chapter0_Contributing.Rmd | 1 + vignettes/chapter1_WhatIsGGIR.Rmd | 7 ++++--- 2 files changed, 5 insertions(+), 3 deletions(-) diff --git a/vignettes/chapter0_Contributing.Rmd b/vignettes/chapter0_Contributing.Rmd index 651d47580..65fa3bd16 100644 --- a/vignettes/chapter0_Contributing.Rmd +++ b/vignettes/chapter0_Contributing.Rmd @@ -32,4 +32,5 @@ You might not have the coding skills to contribute to the code base of GGIR, and - **Apply for funding** to support the development and maintenance of GGIR. GGIR is free software by which we entirely depend on users applying for funding to sponsor our efforts. Funding could be used to support the development of new functionalities, to support improvement of the existing GGIR software code, or to support development of better open-access training materials such as instruction videos. - **Report your issues and questions** in the [GGIR google group](https://groups.google.com/g/RpackageGGIR). - **Proofread the GGIR documentation** and inform us if you miss something or if you found it difficult to follow. +- **Take independent initiatives to complement our efforts**. For example, Prof. Stuart Fairclough created a [series of GGIR video tutorials](https://www.youtube.com/watch?v=AbHgJYTyuiA), Wei Guo and colleagues created the R package [mMARCH.AC](https://doi.org/10.32614/CRAN.package.mMARCH.AC) to post-process GGIR output as described in the supporting [journal article](https://doi.org/10.1123/jmpb.2022-0018), and maybe there are other initiatives we are not aware of. Please communicate about your initiatives via the [GGIR google group](https://groups.google.com/g/rpackageggir/) and/or with GGIR maintainer [Vincent van Hees](https://www.accelting.com/) to avoid duplicated work and to support each other where needed. diff --git a/vignettes/chapter1_WhatIsGGIR.Rmd b/vignettes/chapter1_WhatIsGGIR.Rmd index 6207e86fd..bd09cb085 100644 --- a/vignettes/chapter1_WhatIsGGIR.Rmd +++ b/vignettes/chapter1_WhatIsGGIR.Rmd @@ -67,7 +67,7 @@ Further, we hope GGIR is of use to those without the financial resources for com The philosophy behind the algorithms as implemented in GGIR is that biomechanical explainable (heuristic or knowledge driven) approaches to measurement in science are preferable over purely data-driven approaches. Only when a knowledge driven approach is unrealistic we can consider a data-driven approach. -The idea of a knowledge driven approach is that in order to advance insight in our field of research, it is essential to have an understanding of the causal relation between the phenomena being observed (e.g. acceleration of one body part), the way the (acceleration) sensor works, what we do with the data produced, and how we interpret the data. For example, we know that body acceleration relates to energy expenditure because of physics and human physiology. The abundance of scientific publications that have reported a positive correlation between accelerometer data and energy expenditure only served to confirm that existing knowledge was correct. +The idea of a knowledge driven approach is that in order to advance insight, it is essential to have an understanding of the causal relation between the phenomena being observed (e.g. acceleration of one body part), the way the (acceleration) sensor works, what we do with the data produced, and how we interpret the data. For example, we know that body acceleration relates to energy expenditure because of physics and human physiology. The abundance of scientific publications that have reported a positive correlation between accelerometer data and energy expenditure only served to confirm that existing knowledge was correct. In contrast, data-driven methods focus on optimal correlation between sensor data and reference labels or values, and are much less concerned with causal associations that are the focus of knowledge driven approaches, as defined above. Identical to how correlation is not necessarily equal to causation in health research, the process of measurement can also be confounded. Some examples: We may see differences in body acceleration patterns that correlate with different activity types or different levels of energy expenditure, but that does not mean that we actually measure those activity types or energy expenditure levels. Ignoring such aspects can easily lead to overestimating the value of an accelerometer for measuring those constructs (activity type, etc) and to underestimate the value of an accelerometer of capturing acceleration as a useful measure of behaviour, if appropriately used and interpreted. @@ -76,7 +76,7 @@ A second problem with data-driven methods is that they heavily depend on the ava We argue that such reliable criterion methods do not exist for physical behaviour measurement: 1. Indirect calorimetry and the indicators of energy metabolism that can be derived from it are unable to account for the activity type specific role of body weight on energy metabolism. This makes it impossible to make a standardised comparison of the energy cost of different activity types across individuals that differ in body weight. See also reflections in [this blog post](https://www.accelting.com/updates/why-does-ggir-facilitate-cut-points/). -2. Polysomnography (PSG) is the standard in sleep research. PSG offers a physiological definition of sleep that is impossible to capture with a movement sensor. Therefore, we are forced to simplify our definition of ‘sleep’ towards a definition that can be captured by a movement sensor. As a result, the act of evaluating an accelerometer on its ability to classify sleep with PSG becomes somewhat meaningless as we already know that we are not measuring the same construct as PSG. +2. Polysomnography (PSG) is the standard in sleep research. PSG offers a physiological definition of sleep that is impossible to capture directly with a movement sensor. Therefore, we are forced to simplify our definition of ‘sleep’ towards a definition that can be captured by a movement sensor. As a result, the act of evaluating an accelerometer on its ability to classify sleep with PSG becomes somewhat meaningless as we already know that we are not measuring the same construct as PSG. 3. Activity types are ambiguous to define given the high number of ways they can be performed. This introduces a fundamental level of uncertainty about the robustness of models outside the datasets and context they were developed in. As a result, it is essential to put strong emphasis on algorithms that have descriptive value on their own regardless of whether they offer a high correlation with supposed criterion methods. @@ -99,5 +99,6 @@ Everything you need to type in your R script is `highlighted like this`. This documentation is not intended as an academic review: We only cite publications to clarify the origin of algorithms and we only discuss what is part of GGIR. + Finally, the first version of this documentation was sponsored by Accelting with the commitment that this will remain available as free open-access documentation. -However, things like this are much easier to maintain as a community: We would be grateful for your help to improve the documentation either by giving feedback, pull requests (for those who know how to do it), or financially. +However, open documentation is much easier to maintain as a community: We would be grateful for your help to improve the documentation either by giving feedback (e.g. via v.vanhees at accelting dot com), pull requests (for those who know how to do it), or financially. For example, it would be great if we had funding for creating high quality complementary info graphics and videos.