Skip to content

Latest commit

 

History

History
100 lines (68 loc) · 6.22 KB

File metadata and controls

100 lines (68 loc) · 6.22 KB

📚 Data Dictionary — ACTT Validation Study

This document describes all data sources and variables in the study.
Processed data are archived at: https://doi.org/10.5281/zenodo.15315735

All primary analyses read from a single file: data/acttrust_validation_full.csv


🗃️ Raw Data Sources

Raw data are stored in raw_data/ and were processed by proprietary device software before export. All subsequent processing and analysis is performed in open-source R.

Source Folder Description Format
ActTrust® actigraphy raw_data/ACTTRUST/ Raw exports from ActTrust® devices (Condor Instruments, São Paulo, Brazil). Worn on the dominant hip and non-dominant wrist simultaneously. Records activity in PIM (Proportional Integration Mode) — a measure of the integral of acceleration above a threshold — sampled at 25 Hz, band-pass filtered 0.5–2.7 Hz, and exported at 1-s epochs via Act Studio software. One file per participant per placement. Proprietary .txt
ActiGraph® GT3X+ actigraphy raw_data/Gt3x/ Raw exports from ActiGraph® GT3X+ devices (ActiGraph Corporation, Pensacola FL). Worn on the dominant hip and non-dominant wrist simultaneously. Records activity as vector magnitude (VM) counts — the Euclidean norm of triaxial acceleration — sampled at 30 Hz, band-pass filtered 0.25–2.5 Hz, and exported at 1-s epochs via ActiLife v6.13.4. One file per participant per placement. .gt3x / .csv
Indirect calorimetry raw_data/ERGO/ Breath-by-breath oxygen uptake (VO₂) measured by a Quark CPET metabolic cart (Cosmed®, Rome, Italy) during each treadmill condition. METs are computed as individual VO₂ / 3.5 mL O₂/kg/min. Values represent the mean from the central 4 minutes of each condition to exclude transition effects. Proprietary Cosmed export

Note on count magnitude differences: ActTrust® PIM counts are systematically higher in absolute value than ActiGraph® GT3X+ vector magnitude counts due to differences in filtering, integration method, and hardware sensitivity. The two devices must not be compared directly in absolute terms — use device-specific regression equations and cut-points from this study.


📐 Processed Dataset

data/acttrust_validation_full.csv

One row per participant × treadmill condition. Activity counts and METs are averaged over the central 4 minutes of each condition.

Identifiers and Metadata

Column Type Units Description
ID character Unique participant identifier
Conditions character Treadmill condition. Values: Rest, 3 km/h, 5 km/h, 7 km/h, 9 km/h, Sit-Stands. Sit-Stands is excluded from all primary analyses.
Sex numeric (raw) / character (recoded) 1 = Female, 2 = Male in the CSV. Recoded to Female / Male on load via dplyr::recode().
BMI numeric kg/m² Body mass index

Metabolic Measurement

Column Type Units Description
METs numeric METs Metabolic equivalent of task from indirect calorimetry. Calculated as individual VO₂ / 3.5 mL O₂/kg/min. Values are the mean over the central 4 minutes of each condition. 1 MET ≈ 3.5 mL O₂/kg/min at rest.

Accelerometry — Activity Counts

Four device × placement combinations stored as separate columns. Both are averaged over the central 4 minutes of each condition.

Column Type Units Description
GT3X+ (hip) numeric counts/min ActiGraph® GT3X+ on dominant hip. Reference level in all regression models. Vector magnitude counts; 30 Hz; band-pass 0.25–2.5 Hz; 1-s epochs; exported from ActiLife v6.13.4.
GT3X+ (wrist) numeric counts/min ActiGraph® GT3X+ on non-dominant wrist. Same hardware and processing as hip placement.
ACTT (hip) numeric counts/min ActTrust® on dominant hip. PIM mode; 25 Hz; band-pass 0.5–2.7 Hz; 1-s epochs; exported from Act Studio.
ACTT (wrist) numeric counts/min ActTrust® on non-dominant wrist. Same hardware and processing as hip placement.

data/S1.csv

Supplementary participant-level data. See Zenodo archive for full description.


🔧 Derived Variables

These variables do not exist in the CSV but are computed in index.qmd.

Variable Created by Description
act pivot_longer() Device × placement label after reshaping to long format. Factor with reference level GT3X+ (hip). Values: GT3X+ (hip), GT3X+ (wrist), ACTT (hip), ACTT (wrist).
act_value pivot_longer() Activity count value corresponding to the act device × placement column.
fit_MET `lm() > fitted()
actual_class cut(METs, ...) Observed PA intensity class from measured METs. Breaks: [0,3), [3,6), [6,9), [9,∞).
predicted_class calc_met_class(fit_MET) Predicted PA intensity class from model-fitted METs, using the same breaks as actual_class.
mean_val Bland-Altman section Mean of fit_MET and METs — the x-axis of Bland-Altman plots.
diff_val Bland-Altman section Difference fit_MET − METs — the y-axis of Bland-Altman plots.

🏃 PA Intensity Classification

Class label MET range Intensity
[0,3) 0 ≤ METs < 3 Light
[3,6) 3 ≤ METs < 6 Moderate
[6,9) 6 ≤ METs < 9 Vigorous
[9,inf) METs ≥ 9 Very Vigorous

👤 Participant Characteristics

Summary statistics reported in Table 1 of the manuscript. Individual-level anthropometric data beyond ID, BMI, and Sex are not present as columns in the CSV but are available in the Zenodo archive.

Characteristic Male (n = 34) Female (n = 22)
Age (years) 28.3 ± 4.6 26.8 ± 5.0
Height (cm) 172 ± 6.2 160 ± 7.3
Weight (kg) 78.2 ± 11.9 59.9 ± 9.4
BMI (kg/m²) 23.1 ± 1.6 26.7 ± 2.5