2.2.2. Data Preparation

Raw scores were used for all measures except for the Vineland-II domain-level scores (ABC, communication, daily living skills, socialization, and motor skills), which were only

available as standardized scores. No subscales were removed for excessive missingness (>15% of data points missing across individuals). The DLD had the least amount of missing data (0.71% missing), followed by the Vineland-II (9.22% missing) and the SIB (9.93% missing). Missing data were imputed using chained random forests via the 'missRanger' R package [35]. Next, the DLD scores were inverted for consistent directionality with the other measures. All analyses were completed in R v 4.0.0 [36] and the significance level set to 0.05.
