34 Chapter 2 duration of the trial. This resulted in gaze data of 746 trials in total (out of 768; 2.9% missing data) of 48 participants (out 79 participants for the fMRI study). Areas of interest (AOI) in the stimuli were created around the left and right eye and the overall face area for all presented videos using MATLABs cascade object detector (Viola & Jones, 2001). This MATLAB toolbox used an established algorithm for face and facial feature detection. More specifically, for each frame of each video, this algorithm outputted rectangular AOIs encompassing the left eye, right eye, and overall face area. The primary eye gaze measure used was the percentage of dwell time within the AOIs (i.e., left eye, right eye, overall face area) per video stimulus, as part of the total duration of the video (16-38 s), in which dwell time is defined as the total amount of time spent looking within an AOI and includes all types of eye movements (e.g., fixations and saccades). The gaze data within the left and right eye AOIs were combined into a single AOI of the eye region for further analyses. To examine whether the amount of eye contact was moderated by gaze direction (2 levels: Direct versus averted gaze) and target identity (4 levels: Own child, unfamiliar child, unfamiliar adult, self), we used a generalized linear mixed regression model in R with gaze direction and target as predictors of the percentage of dwell time within the eye region of the targets. Furthermore, we examined whether parents who looked more at the eyes of targets reported a higher mood and enhanced feelings of connectedness. To do this, we used a similar generalized linear mixed regression model and included the percentage of dwell time within the eye region, gaze direction, and target as predictors of self-reported affect. We controlled for gender of the parents and current MDD/dysthymia diagnosis of the child of the parent in all analyses including affective and gaze responses of parents. In two cases, two parents of the same adolescent participated in the task. To control for potential dependencies in the data we added a covariate to the model indicating whether parents were part of the same family. Significance was set at p <.05 (two-tailed) and Cohen’s d effect sizes were calculated for significant effects. Neuroimaging analyses MRI data were preprocessed and analyzed using SPM12 (Wellcome Trust Centre for Neuroimaging, University College London). Functional MR images were slice-time corrected, corrected for field-strength inhomogeneity’s using b0 field maps, unwarped and realigned, coregistered to subject-specific structural images, normalized to MNI space using the DARTEL toolbox (Ashburner, 2007), and smoothed using an 8-mm full width at half maximum isotropic Gaussian kernel. Raw and preprocessed data were checked for quality, registration, and movement. Head movement did not exceed 1 voxel (i.e., 3 mm) for any of the participants (M = 0.09 mm, SD = 0.05 mm, range: 0.002 – 2.759 mm). Furthermore, we corrected for serial autocorrelations using a first order autoregressive model (AR(1)). We removed low-frequency
RkJQdWJsaXNoZXIy MjY0ODMw