2.4 Statistical analyses
Initially, an ANOVA and chi-square analysis were done to compare demographic data. Then, MANCOVA was used to compare the NSS and neurocognitive performance with age as a covariable. The Fisher’s least significant difference (LSD) method was used to conduct post hoc comparisons between EBP and HC, EBP and FDR, and FDR and HC. Each test was subjected to Bonferroni corrections to account for multiple testing. In addition, Using NSS and neurocognitive performance as predictor variables, a step-by-step linear discriminant functional analysis was performed to predict group membership (EBP, FDR, and HC) (EBP and HC) (FDR and HC) (EBP and FDR). A two-tailed p value of less than 0.05 was considered statistically significant. All the statistical analyses were carried out with SPSS22.
3 results
3.1 Demographic data
There were no significant differences between the three groups in terms of gender, handiness, or education level. In RP, the median number of episodes was 3.09±0.845, the median HAMD score was 3.58±1.439, and the median YMRS score was 2.89±1.107. There were significant differences in age across the three groups (Table 1).
3.2 Neurological soft signsThe results revealed that the EBP had significantly more anomalies in the motor coordination subitem and total NSS score than the FDR (p<0.001; Bonferroni-corrected p<0.01). These subitems were also significantly higher in the EBP compared to the healthy controls, with Bonferroni correction accounting for all significant differences. The FDR scored significantly higher on motor coordination subitems (p<0.001; Bonferroni-corrected p<0.01) and total NSS (p=0.001; Bonferroni-corrected p<0.05) than HC. Furthermore, EBP (p<0.001; Bonferroni-corrected p<0.01) showed larger sensory integration abnormalities than FDR and healthy controls, however, no significant differences were found between FDR and HC. Finally, there were no significant differences between the three groups on the disinhibition items (Table 2).3.3 Neurocognitive performanceExcept for RS, there were significant differences in the six MCCB subscales across the three groups(p<0.001).RP and FDR performed significantly worse than healthy controls in all cognitive areas except RS. EBP fared much worse on IPS, VL, and WM than FDR. Furthermore, all significant differences were subjected to the Bonferroni correction (Table 3).3.4 Discriminant functional analysisThe Wilks method’s F values, which serve as the standards for determining the inclusion and exclusion of predictive factors, are 3.84 and 2.71, respectively. Only 65.4% of the original 243 samples were properly classified by stepwise linear discriminant function analysis when the three groups were mixed when NSS (motor coordination) and neurocognitive performance (IPS, VL, WM) were employed as predictors (EBP 64.2%, FDR 81.5%, HC 50.6 %). The discriminant function properly identified 88.3%of the original sample when EBP and HC were merged ((EBP 87.7 % and HC 88.9 %). When FDR and HC were intermingled, the discriminant function accurately identified 81.5% of the original sample (EBP 84% and HC 68%). When EBP and FDR were mixed, the discriminant function accurately classified 62.3%of the original sample (EBP 58% and 66.7%HC). When the NSS or neurocognitive performance were entered separately into the discriminant function analysis, the accuracy of categorizing each group was lower than when both were integrated as predictor factors. (Table 4).