Inoc Visual Index
In Playing the Archive we sonified data that was collected immediately after a child received an inoculation (stuck with a needle). The interest was to convert the mother-infant (dyad) exchange into a musical form. Check out the performance. We continue to work on ways to present the data visually – to support data analysis and discovery of additional musical forms. As part of The Emotional Beginnings Project, Cindy Stifter and her team observed and video recorded mothers and infants during their routine inoculations at the doctor’s office. Working from the videotapes, the intensity of the infant’s distress (crying) and the actions of the parent soothing the child were coded into data. Each observed dyad is represented as a multivariate time-series of categorical data, which may be explored for patterns in the mother-infant behavior and the regulation strategies they employ as they recover from this generally distressing scenario. Each dyad is unique – but can also be compared to the other dyads.
Along side our statistical modeling of these data, we are exploring how the data and patterns can be represented visually through our use of symbols, coloration, and change over time. Like musical scores we can “read” the sounds of the child and actions of the mother from the inoculation shot moment (left edge – RED) until they achieve an extended period of no-distress or until a designated length of time has elapsed (right edge). In the example shown below, changes in the intensity of one infant’s distress are represented by the color saturation and the amount of vertical space that is “color-occupied”; The circles indicates the presence of parent soothing behaviors at a given moment of observed time (5 second intervals). If behavior occurs continuously then a horizontal, centered line connects the observed and coded intervals.
The dyad plots inspire and are inspired by interests that seek to understand within-dyad change in the inoculation context and in relation to a particular stimulus – the needle prick. But these data were collected as part of a larger sample where between-dyad differences are also a focus of modeling and inference. So these data, obtained through naturalistic observation, include data from ~150 dyads. A related set of visualizations are used to represent the between-dyad variation in infant distress, parent behavior, and total duration of task (e.g., <1min, up to 4 min). The sample index plot of infant distress below affords a view of the entire sample – with each dyad represented by a single row – that highlights differences in the task duration and level of crying. General and immediate impressions emerge with respect to observed time in the recorded task, where some cases provide “short” 1-minute segments, others “long” 4-minute segments, and others in-between. A mostly solid and bright pink “start” section indicates successful (from an experimental perspective) elicitation of distress from which self-regulation and parent soothing strategies may be studied. The shift from darker red on the left to lighter shades of pink on the right depicts how dyads regulated the early distress towards (mostly) no-distress (filled black regions). The heterogeneity across dyads is very apparent.
Like the infant distress sample-level plot, 10 specific behaviors the parents used in their attempts to calm the child are visualized separately as index plots. The presence (e.g., green-filled regions) or absence (e.g., black-filled regions) of the each behavior is shown. Again, sample-level and dyad-level relative frequencies may be observed. For example, rocking behavior (overall high frequency; green) and use of a pacifier (overall low frequency; blue) are shown below.
To look at how the parents’ behaviors are deployed in tandem or triples etc, we can look at all the 10 panels simultaneously.
A major focus of our analytical innovations revolves around the connections between dyad-level and sample-level analysis – differences in scale. “Zooming” in to the infant distress index below shows where we focus to the specific “row” for the dyad plot presented above. And doing the same (position) for the rest of the discrete parent behaviors (in panel) shows how the dyad-level and sample-level data are connected.
This zooming in and out from the sample-level index plots to the dyad-level plots provides a visual accompaniment that inspires and informs our integration of between-dyad and within-dyad analyses. A first run at those analyses explores how the structure of parent-child behaviors differs between needle pricks received at 2 months of age and at 6 months of age (Stifter et al., submitted; Rovine et al., 2010).