Please excuse me while I nerd out all over your computer screen. I recently turned a corner on my appreciation of the value of quantitative social science, having taken a structural equation modelling class last winter, and today I’m going to share a little of that with you. While I’m still a qualitative researcher through and through, this course taught me that there is great value in understanding how scales are constructed and what that means about how we can interpret results from survey-takers.
What, you might ask, does any of this have to do with eating disorders? Plenty. A while back, Shiran wrote a post about the issues with the Eating Disorders Examination Questionnaire. Her post didn’t focus on the scale psychometrics – that is, how well the scale measures what it is supposed to measure and how consistent it is – but still reveals how questionnaires used to determine whether or not someone can be diagnosed with an eating disorder are not perfect.
Think about it this way: you’re filling out a questionnaire as you sit in the cold waiting room to an eating disorder clinic, having filled out 10 other questionnaires, maybe having driven an hour to get to the clinic, wrestling in your mind about whether you’re really ready for treatment, perhaps feeling twitchy or headachey or tired. Does your response on a survey questionnaire represent your real state of being forever? I would argue that while some scales are better than others at accounting for the individual variability of our subjectivities each day, your responses are not “objective reality.” Rather, they comprise your interpretation of the question, the space you’re in (physically and mentally) that day, where the question is located on the scale, and the rest of your context (e.g., your social location, or ethnicity, gender, sexuality, ability, etc.).
In this post, I’ll look at a 2016 article by Hagan, Forbush & Chen, in which they explore the construct of dietary restraint (a common eating disorders subscale) and whether it is “a unitary or multi-faceted construct” – that is, whether there is a single “thing” that constitutes dietary restraint, or whether dietary restraint is made up of a few sub-“things.”
The authors situate their study by noting the significant number of US adults who are currently dieting – upwards of 108 million, if you take self-report figures as valid. The diet industry is big business, they note – to the tune of $61 billion per year. Further, not only do we know that diets are rarely “effective” (taking maintained weight loss to be “effective”) but we also know that some people who diet may develop eating disorders (see Neumark-Sztainer et al., 2006).
Hagan, Forbush & Chen introduce readers to the murky terrain of evidence around the role of dietary restraint in binge eating episodes. They cite evidence both for and against the hypothesis that dietary restraint leads to increased binge eating and, ultimately, weight gain. Because of the conflicted nature of the evidence, the authors wondered whether discrepancies in findings might be partially explained by the fact that these studies might be measuring different things – in other words, researchers might be understanding and measuring dietary restraint in different ways, leading to conflicting results.
To find out, they recruited 271 women and 62 men from an undergraduate education setting and 193 women and 214 men from the community to fill out a number of questionnaires designed to measure eating behaviours and restraint behaviours. These were:
- The Dutch Eating Behaviour Questionnaire
- The Eating Disorder Examination Questionnaire
- The Eating Disorder Inventory
- The Eating Pathology Symptoms Inventory
- The Restraint Scale
- The Inventory of Depressive and Anxiety Symptoms
- The Three-Factor Eating Questionaire
- The Frost Multidimensional Perfectionism Scale
The conducted structural equation modelling techniques on the answers (both exploratory and confirmatory analyses). To briefly explain what that means, these statistical techniques are designed to investigate which items in a questionnaire go together by looking for response patterns. Once they’d established a basic understanding of the different elements of the overall concept (dietary restraint), they looked to confirm their results. In the confirmatory phase, they looked at whether men and women answered the questionnaires in a similar way (i.e., if they seem to have interpreted the information in the same way – this means looking for patterns in responses) as well as whether community and undergraduate samples responded similarly.
After the first phase of analysis, the authors determined that dietary restraint might be made up of three different parts:
- Calorie Counting
- Preoccupation with Dieting
- Weight-Focused Restraint
Once they had confirmed that their model fit and that both men and women and community and university samples answered in a similarly patterned way, the authors looked at which responses were correlated with risk for eating disorders. They found that all of the factors were linked to increased risk for eating disorders; preoccupation with dieting and weight-focused restraint were both associated with having a higher BMI.
Why Does it Matter?
Overall, I think this study is interesting and relevant. I do disagree with how the authors frame much of their discussion section, however, and wish to highlight how societal ideology works its way into even quantitative social science by drawing out some of my points of disagreement.
What matters about this study is that:
- It shows us that the scales we use to measure things are not perfect. When scales attempt to lump the multiple parts of “dietary restraint” into one, they might miss variations. For instance, they might find that some people are endorsing dietary restraint when they mean counting calories, while others mean focusing on their weight.
- It begins to support the case that societally reinforcing a focus on calorie counting and other aspects of dietary restraint might be harmful insofar as they may increase the risk of eating disorders.
The authors suggest that both eating disorder and weight-loss interventions can benefit from the information, by promoting self-regulation and encouraging clients to avoid labelling foods as off limits and not focus on the thin ideal.
However, how tenable is this suggestion in a society that actively encourages the reverse? The expectation, here, is that clients will be able to develop enough inner resilience to avoid the sway of societal messaging. Further, by insinuating that this is a relevant finding because it will help people lose weight, the expectation that weight loss is desirable still dances between the lines of the text.
I think it’s important to note that I am in no way blaming the authors for this interpretation. Of course the argument is framed in this way. It’s worth noting that in order to have research funded and articles published in the eating disorders field, there’s an expectation that researchers will acknowledge the ‘but what about the obesity epidemic” question.
What I think is worth drawing out is just this – how research on things like dietary restraint are only societally valued if they tell us how we can better encourage weight loss. This leads to a reproduction of some of the very contradictions that leave us feeling bad about our bodies – and then feeling bad about feeling bad about our bodies because this is bad for our health. How is it possible to simultaneously discourage a focus on thinness while using this information to encourage people to aim for thinness?
I yearn for a world in which, in order to do research relevant to eating disorders, we don’t need to look for how our results might also lead to a thinner population. The inherent tension there is head-scratching for me. Why don’t we focus more on changing society to be more welcoming of diverse bodies, rather than trying to change people to fit societal norms, which are (even by their own admission, at times) never attainable?
Hagan KE, Forbush KT, & Chen PY (2016). Is Dietary Restraint a Unitary or Multi-Faceted Construct? Psychological assessment PMID: 27991825