Autism Quotient (AQ; Baron-Cohen et al., 2001)
The AQ is a 50-item self-report questionnaire assessing autistic traits
in adults. Participants respond to statements such as “I am
fascinated by dates” and “I find social situations easy” by
indicating if they “definitely agree”, “slightly agree”,
slightly disagree” or “definitely disagree”. Around half of
the items are reverse scored, with responses indicating the presence of
an autistic trait scoring 1, therefore scores range from 0-50 with
scores of 32 and above indicative of the likely presence of ASC. The AQ
has good test-retest reliability (r =.70, Baron-Cohen et al.,
2001) and has been used to measure autistic traits in the general
population as well as clinical samples. Within the AQ, items relate to
five subtypes of autistic traits; social skills, attention switching,
attention to detail, communication and imagination.
Statistical analysis
Results were analysed using SPSS 24, with descriptive procedures run to
determine means and standard deviations, followed by Pearson
correlations to reveal intercorrelations between measures as well as to
determine the proportion of participants scoring above the clinical
cut-off on measures where possible. An additional t-test was conducted
to determine if the mean AQ score of the sample was significantly
different to that reported elsewhere for the general population.
Mediation analysis allowed examination of indirect relationships between
anxiety, ASC and associated symptoms such as IU and SOR. The mediating
effects of SOR and IU were investigated using the Baron and Kenny (1986)
four step method of mediation. Where Step Four results in both
predictors remaining significant, a Sobel test can be conducted to
determine whether the reduction in the effect of the IV was significant,
therefore indicating a significant mediation effect.
Sensitivity analysis and Missing Data
Conservative a priori sensitivity analysis conducted using an online
statistics calculator (Soper, 2019) demonstrated that multiple
regression with a minimum sample size of 97 would achieve a power of 0.8
(α=.005) with at least medium effect sizes f2=.15.
Regarding missing data, participants with missing data for more than two
items in each questionnaire were removed before the analysis. Of those
remaining, the average score for items on the corresponding
questionnaire was used to fill in missing data for all measures apart
from the AQ, where a missing item was assumed to have scored 0 due to
the binary nature of the scoring system.
Results
Descriptive Statistics
Means, standard deviations and intercorrelations for each measure are
displayed in Table 1, as well as the proportion of participants who
scored on or above clinical cut-off where possible. All measures were
significantly correlated with each other at p <.001. The
mean AQ score for this sample (M =25.18, SD =8.21) was
significantly higher than the average female score in the general
population used in the original paper (M =15.40, SD =5.70;
Baron-Cohen et al., 2001), t (359)=22.59, p <.00.
Additionally, 76 participants (21.1%) scored on or above the clinical
cut-off on the AQ, indicating the likely presence of ASC, compared to
1% of females in the general population (Baron-Cohen et al., 2001).