*2.3. Protocol*

The parents signed an informed consent form and the child gave verbal consent to participate in the trial. The parents filled out a demographic questionnaire. Then, two sensors were attached to the wrists of the child and two to the ankles using elastic belts. The child was asked to ambulate with the sensors and report any inconvenience. Then, the child was asked to perform at least 8 consecutive jumping jacks (also called star jumps). The jumps were demonstrated before data recording began. This was performed twice: once at a self-selected rhythm and once guided by the metronome, located 1 m behind the subject.

#### *2.4. Post Analysis*

The data from each accelerometer were downloaded to a personal computer and then exported as CSV files using Actilife™ software version 5.10.0 (ActiGraph, Pensacola, FL, USA). A custom code was created in LabView (v2015, National Instruments, Austin, TX, USA). The acceleration magnitude was calculated using the accelerations of the 3 axes, *Ax*, *Ay*, and *Az*, as:

$$\text{Limb acceleration} \left[ \frac{m}{s^2} \right] = \sqrt{A\_x^2 + A\_y^2 + A\_z^2} \tag{1}$$

Five consecutive jumps were taken (the first and last jumps were excluded from the analysis as initiation and termination of movement). We designed this protocol to include a small number of jumps because a longer jumping sequence might involve fatigue, which will produce bias in the results of limb accelerations. Also, the cooperation levels and concentration span of 5-year-olds in the clinical settings might be low, so that a longer examination might not be possible. The maximal acceleration of each limb of each jump was calculated. Also, the timing of the peak acceleration (in seconds) of each jump was calculated and the following inter-limb timing measures were computed using the following formulas:

$$TL\,\left[\text{sec}\right] = \frac{\sum \left(t\_L - t\_R\right)}{5} \tag{2}$$

$$LV\,\left[\sec\right] = \sqrt{\frac{\sum TL^2 - \frac{\left(\sum TL\right)^2}{5}}{5}}\tag{3}$$

The timing lag (*TL*) is the difference in the timing of the maximal acceleration of the left lower or upper limb (*tL*) and right lower or upper limb (*tR*), averaged for five consecutive jumps. It is defined similarly as constant error of two measures (left and right limb herein) that are expected to be identical during symmetric movement [19,20]. A positive TL value denotes that the left limb reached maximal acceleration sooner compared to the right limb. As the TL value decreases towards zero, the coordination of the two limbs in reaching maximal acceleration is higher, meaning that they are more in-phase. The lag variability (*LV*) is the variation of TL across the 5 jumps. It is defined in a similar manner as variable error of two measures and produces the average of the standard deviation [19]. High LV is indicative of low consistency between jumps. These measures are calculated separately for each condition, with and without the metronome rhythm, and presented in seconds.

All of the statistical analyses were performed in IBM SPSS Statistics 25. Mann–Whitney U test was used to test for between-group differences in age and body mass index (BMI), and the Chi-square test was used to test for between-group differences in sex. ANOVA analysis of group (ASD and TD) × condition (with and without metronome) was performed. Post hoc tests were administered according to the findings. Effect size estimates, *r*, for Mann–Whitney non-parametric tests were calculated according to [21]:

$$r = \frac{Z}{\sqrt{N}}\tag{4}$$

Statistical significance was set to *p* < 0.05.
