5.4.1. Activities

As depicted in the Table 3, all the participants spent the majority of their time in a sitting posture ( *M* = 79.5, *SD* = 6.1). Designing PCBs, coding, soldering, debugging firmware/software, reading/writing papers, meetings, writing emails, having coffee/lunch and watching YouTube, were the major activities that participants reported doing during the study.


**Table 3.** Percentage of being in sitting posture of each participant during the 8 h field study.

Most of the participants marked higher stress levels (>4 on a 7-pnt Likert scale) when they performed office work-related activities, such as coding, debugging and writing papers. A summary of the self-reported stress level, level of energy and level of pleasantness across all participants is depicted in Table 4. According to the table, when participants felt stressed (Stress Level > 4), a lower level of pleasantness ( *M* = 3.8, *SD* = 1.4) was reported compared to when participants were relaxed (Stress Level < 4). Welch's *t*-test further confirmed that there is a significant difference (*t* = −4.10; *p* < 0.05). Moreover, Welch's *t*-test showed a significantly higher task load when participants were stressed, compared to being relaxed (*t* = 7.02; *p* < 0.05). These findings confirm that we have collected two significantly different levels of stress (stress and relaxation).


**Table 4.** Summary of level of pleasantness, energy level and task load.

Moreover, the rated "level of energy" was not statistically different when dividing the gathered data into two groups (stressed and relaxed) following the Welch's *t*-test (*t* = 1.64; *p* > 0.05). This means the perceived level of energy does not relate to stress since energy can be both positive and negative.

#### 5.4.2. Electrodermal Activity (EDA)

Table 5 summarises the average EDA slopes of the participants in stressed and relaxed conditions. When a participant marked the previous hour with a stress level > 4, we considered that hour as a high stress duration. We considered ratings below 4 as a relaxed period. When stressed, the majority of participants (5 out of 6) demonstrated an average positive EDA slope. Comparatively, when participants were relaxed, 8 out of 10 participants showed an overall negative EDA slope. However, none of the participants showed a significantly different mean EDA slope according to Welch's *t*-test (see Table 5). The limited difference is due to high variances of EDA readings caused by external factors, such as motion artefacts, ambient temperature variances and loose contact of electrodes on the skin.


**Table 5.** Summary of EDA slopes of each participants while being stressed and relaxed.

#### 5.4.3. Overall in Field Validation

We now analyse how our model compares to the users' Self-Reported Stress Level (SRSL). Figure 7 depicts the SRSL and *RS* of each hour for each participant. The graphs already demonstrate a positive correlation. To identify the significance of this relationship, we calculated the Pearson's Correlation Coefficients (*r*) for each participant. All participants showed a positive correlation coefficient, with an average of *r* = 0.79 (*SD* = 0.10). Eight out of 10 participants demonstrated a statistically significant positive relationship (*r* > 0.7, *p* < 0.05) between SRSL and *RS*. Among these participants, three had a high correlation coefficient (*r* > 0.9) which shows strong relationship between SRSL and *RS*. Only two participants showed a moderate Pearson's Correlation Coefficient (*r* = 0.67 and *r* = 0.65). However, this result was not statistically significant (*p* > 0.05). Even from those two participants, P8 reported lower stress levels (<4) for the entire study duration, which is also confirmed by our model showing a lower *RS* (<0.5) value during the entire study period. Overall, we can conclude that all the participants

showed a positive correlation, given the majority demonstrated a significantly high correlation between SRSL and *RS*. Such findings evidences the robustness of our model in-field, beyond a controlled laboratory setting.

**Figure 7.** The figure shows the Self-Reported Stress Level (*SRSL*) and *RS* for each participant for each hour. Where *R*S = (*nWindows*sitting(*Stressed*)/*nWindows*sitting(*Total*)). The Pearson's Correlation Coefficient *r* and the level of significance *p* for each participant's *SRSL* and *RS*is also depicted.
