Table 11 Methods for data quality assessment

Eye-tracking

 - VOF task

Data quality was manually inspected by looking at the amount of gaze data points present (X- and Y-coordinates of at least one eye) and the amount of gaze data points directed to a stimulus (in contrast to undirected gaze data points). Data was considered of acceptable quality when post-calibration was possible. For informative purpose, data loss and accuracy were reported. Criteria to determine if a stimulus has been seen and to determine reaction time can be found in Kooiker et al. [16]

 - Social preference task

A slide was considered valid to assess attention capture (first looks) to faces when participants were fixated at the center of the screen in the last 500 ms before trial onset and fixated on an Area Of Interest (AOI) in the first three seconds after trial onset. A trial was considered valid to assess sustained attention (total fixation duration) to faces when the participant spent longer than one second fixating AOIs on the slide (based on criteria from previous literature using the same task) [19, 37, 51, 52]. Acceptable data quality was defined as having at least three valid trials for either the attention capture or sustained attention analyses. Data loss (%), precision (noise level), and accuracy were reported for informative purpose

fNIRS

Data quality was visually inspected and graded per channel per participant. In addition, the signal quality index (SQI)—an algorithm for quantitative assessment of fNIRS signal quality—was calculated per channel. Both the manual grades and SQI ranged from zero (very bad quality) to five (very good quality). Participants were considered to have acceptable data quality when the average of the visually inspected grade and SQI was 3, 5 or higher [53]

Indirect calorimetry

Data quality was considered acceptable if a steady-state ventilation of at least 5 min was reached, and when minute-to-minute oxygen consumption and carbon dioxide elimination varied by ≤ 10%. The first 5 min of a measurement were not taken into account, because the RMR is known to decline and then stabilize due to acclimatization of the participant. These criteria were based on previous literature [54,55,56]. In addition, data quality was considered unacceptable if the result was clinically impossible (e.g., a difference between measured and predicted RMR of 60%)

BIA

Data were considered of unacceptable quality when the BIA showed the notification “Perform the measurement again. Incorrect posture may cause inaccurate results,” which was displayed in case two or more impedance values were reversed, in case of high impedance values in the trunk, and in case of big impedance drops. This can be caused by movement of the child, when the segments are not separated from each other (e.g., arms touched the torso), or when the parent touched the child. In addition, the quality was considered unacceptable when the outcome was clinically impossible (e.g., extremely low fat percentage in combination with high BMI)

BOD POD

Data were considered of unacceptable quality when the outcome was clinically impossible, for example, extremely low fat levels in combination with a high BMI. Further quality measures were not available for the BOD POD