Erratum: Weight Perception, Weight Stigma Concerns, and Overeating

Obesity (2018) 26, 1312–1321. doi:https://doi.org/10.1002/oby.22224

The authors of this article recently analyzed data reported in this publication in order to inform a new project, and in accessing the data files, they found some discrepancies in the data file which were caused by researcher error.

1. Because of an error in the coding of participants who completed the questionnaires in full and without failing attention checks, 2 participants from Study 1 were incorrectly excluded from analysis and 91 participants were incorrectly included in analyses in Study 2. The authors have rerun the analyses and there were no changes to any of the results, aside from very minor deviations to some values (e.g., percentage of effect explained changing from 23.3% to 22.9%). All tests of statistical significance remain significant and nonsignificant as in the published manuscript. The statistical size of effects also remain the same.

2. In Study 1, participants reported on their annual income. Response options included: less than $25,000 (student) and less than $25,000 (nonstudent). They were mistakenly treated as separate response options in the analyses that used income as an ordinal variable, when in reality one was neither larger or smaller than the other and they should have been combined to form a single ordinal category of less than $25,000. This error was in part caused by there being subtle differences in the way that variables used in both Study 1 and Study 2 were measured. The authors have rerun the analyses and there were no changes to any results, aside from very minor deviations to some values. All tests of statistical significance remain significant and nonsignificant as in the published manuscript. The statistical size of effects also remain the same.

3. In addition to the income variable, on close inspection, authors have identified that although there were minor differences (e.g., reporting of use of 7‐point rather than 5‐point Likert scale response format) between the way that the variables; education level, weight stigma concerns, depressive symptoms, and perceived weight discrimination were measured in Study 2 compared with Study 1, this was not made explicit in the published supplementary materials.

4. Authors also identified a typo in the footnote of Tables 2 and 3.

Table 2.

Indirect effect of perceived overweight on stress‐induced overeating via weight stigma concerns (Study 1)

		Unstandard.coeff.	SE	P	Bootstrap 95% CI	Model R2 / proportion mediated (%)	Stand. coeff. a	SE	95% CI
Model 1 b	Path A	3.71	0.59	<0.001	[2.56, 4.87]	–	0.57	0.09	[0.39, 0.75]
	Path B	0.04	0.01	<0.001	[0.02, 0.05]	–	0.27	0.05	[0.17, 0.38]
	Indirect effect	0.14	0.03	–	[0.08, 0.22]	31.7%	0.16	0.04	[0.09, 0.24]
	Path C (total effect)	0.44	0.09	<0.001	[0.27, 0.61]	0.166	0.49	0.10	[0.30, 0.69]
	Path C’ (direct effect)	0.30	0.09	<0.001	[0.13, 0.48]	0.215	0.34	0.10	[0.14, 0.53]
Model 2 c	Path A	2.82	0.57	<0.001	[1.70, 3.94]	–	0.43	0.09	[0.26, 0.60]
	Path B	0.03	0.01	<0.001	[0.01, 0.04]	–	0.21	0.06	[0.10, 0.32]
	Indirect effect	0.08	0.03	–	[0.04, 0.14]	22.9%	0.09	0.03	[0.04, 0.16]
	Path C (total effect)	0.35	0.09	<0.001	[0.18, 0.53]	0.229	0.40	0.10	[0.20, 0.59]
	Path C’ (direct effect)	0.27	0.09	0.003	[0.10, 0.45]	0.252	0.30	0.10	[0.11, 0.50]

Path A = correlation between perceived overweight and weight stigma concerns; Path B = correlation between weight stigma concerns and stress‐induced overeating; indirect effect = effect of perceived overweight on stress‐induced overeating through weight stigma concerns; Path C = effect of perceived overweight on stress‐induced overeating when weight stigma concerns is not present in the model; Path C’ = correlation between perceived overweight and stress‐induced overeating after taking weight stigma concerns into account.

^aCalculated by repeating the analysis of indirect effects on z scores for any continuous variables (e.g. age, BMI, weight stigma concerns, stress‐induced overeating, neuroticism, perceived weight discrimination, depression).

^bAdjusted for age, gender, ethnicity (white, nonwhite), income, education, chronic illness, and BMI.

^cAdjusted for variables listed for Model 1, plus neuroticism, perceived weight discrimination, and depression.

Table 3.

Indirect effect of perceived overweight on uncontrolled eating via weight stigma concerns (Study 2)

		Unstandard. Coeff.	SE	P	Bootstrap 95% CI	Model R2 / proportion mediated (%)	Stand. Coeff. a	SE	95% CI
Model 1 b	Path A	6.62	0.77	<0.001	[5.10, 8.13]	–	0.73	0.08	[0.56, 0.89]
	Path B	1.03	0.11	<0.001	[0.82, 1.24]	–	0.41	0.04	[0.33, 0.49]
	Indirect effect	6.80	1.07	–	[4.85, 9.06]	60.3%	0.30	0.05	[0.22, 0.40]
	Path C (total effect)	11.27	2.21	<0.001	[6.93, 15.61]	0.133	0.49	0.10	[0.30, 0.69]
	Path C’ (direct effect)	4.47	2.16	0.04	[0.23, 8.71]	0.249	0.20	0.09	[0.01, 0.38]
Model 2 c	Path A	2.91	0.72	<0.001	[1.50, 4.33]	–	0.32	0.08	[0.16, 0.48]
	Path B	0.51	0.14	<0.001	[0.23, 0.79]	–	0.20	0.06	[0.09, 0.31]
	Indirect effect	1.49	0.55	–	[0.65, 2.84]	47.1%	0.07	0.02	[0.03, 0.13]
	Path C (total effect)	3.16	2.20	0.15	[−1.16, 7.49]	0.295	0.14	0.10	[−0.05, 0.33]
	Path C’ (direct effect)	1.67	2.19	0.45	[−2.63, 5.98]	0.312	0.07	0.10	[−0.12, 0.26]

Path A = correlation between perceived overweight and weight stigma concerns; Path B = correlation between weight stigma concerns and uncontrolled eating; indirect effect = effect of perceived overweight on uncontrolled eating through weight stigma concerns; Path C = effect of perceived overweight on uncontrolled eating when weight stigma concerns is not present in the model; Path C’ = correlation between perceived overweight and uncontrolled eating after taking weight stigma concerns into account.

^aCalculated by repeating the analysis of indirect effects on z scores for any continuous variables (e.g. age, BMI, weight stigma concerns, uncontrolled eating, neuroticism, perceived weight discrimination, depression, self‐esteem, body dissatisfaction, physical activity).

^bAdjusted for variables listed for Study 1, Model 1.

^cAdjusted for variables listed for Study 1, Model 2, plus self‐esteem, body dissatisfaction, physical activity.

Table 1.

Sample characteristics

	Study 1, mean (SD)/%		Study 2, mean (SD)/%
Age	36.7 (12.0)	Age	38.8 (12.2)
BMI	27.1 (6.1)	BMI	26.9 (5.6)
Gender (% women)	57.8	Gender (% women)	57.6
Perceived overweight	61.8	Perceived overweight	58.7
Long‐standing illness (% yes)	24.0	Long‐standing illness (% yes)	23.1
Ethnicity		Ethnicity
White	78.4	White	81.5
Black	7.3	Black	6.9
Asian	5.5	Asian	4.8
Hispanic	5.3	Hispanic	4.0
Mixed	2.8	Mixed	2.2
Other	0.6	Other	0.6
Annual income		Annual income
Less than $25,000	33.5	Less than $26,000	28.5
Between $25,000 and $39,999	20.8	Between $26,000 and $39,999	20.4
Between $40,000 and $49,999	14.9	Between $40,000 and $49,999	13.4
Between $50,000 and $74,999	19.2	Between $50,000 and $74,999	22.7
Between $75,000 and $99,999	6.9	Between $75,000 and $99,999	8.7
$100,000 or higher	4.8	$100,000 or higher	6.2
Educational attainment		Educational attainment
Never completed high school	0.2	Never completed high school	0.3
Completed high school	17.4	Completed high school	36.8
College	23.8	Bachelor's degree	48.8
Bachelor's degree	43.0	Master's degree	10.1
Master's degree	12.3	PhD/professional degree	4.0
PhD/professional degree	3.4

As a results of these issues, the following changes are needed:

Abstract

• Number of participants changed to 1,147.

• Results section of Abstract should say, “This explained 22.9% (Study 1) and 60.3% (Study 2) of the variance in the relationship between perceived overweight and overeating tendencies.

Methods, Study 1

• Editing of exclusion criteria changes text to read, “98 participants were excluded for not completing the questionnaire in full, 72 participants failed at least one attention check, 33 reported a self‐perception of underweight, and 10 reported implausible BMIs. This resulted in a final analytic sample of 505 participants.”

• Likert scale description should be changed to 5‐point (from 7‐point) and five discriminatory experience items instead of six.

• Chronbach α values changed to 0.96 (weight stigma concerns), 0.83 (neuroticism measure), and 0.87 (perceived weight discrimination) from, respectively, 0.95, 0.80, and 0.89.

Results, Study 1

Text should read, “In our first model (Table 2, Model 1), weight perception was a significant predictor of weight stigma concerns (unstandardized coefficient, B = 3.71; SE = 0.59; P < 0.001), and in turn, weight stigma concerns were a significant predictor of stress‐induced overeating (B = 0.04; SE = 0.01; P < 0.001). Perceived overweight (relative to perceived normal weight) had a significant indirect effect on stress‐induced overeating via weight stigma concerns (bootstrap estimate = 0.14; SE = 0.03; 95% CI: 0.08 to 0.22), with weight stigma concerns explaining 31.7% of the variance in the relationship between perceived overweight and stress‐induced overeating. In the fully adjusted model (Table 2, Model 2), perceived overweight relative to perceived normal weight had a significant indirect effect on stress‐induced overeating via weight stigma concerns (bootstrap estimate = 0.08; SE = 0.03; 95% CI: 0.04 to 0.14), and weight stigma concerns explained 22.9% of the variance in the relationship between perceived overweight and stress‐induced overeating. For gender, the index of moderated mediation was not significant (bootstrap estimate = 0.06; SE = 0.04; 95% CI: −0.03 to 0.15), suggesting that gender did not moderate the indirect effect of perceived overweight on stress‐induced overeating via weight stigma concerns. For weight perception accuracy the index of moderated mediation was not significant (bootstrap estimate = −0.02; SE = 0.07; 95% CI: −0.18 to 0.11), indicating that weight perception accuracy did not moderate the indirect effect of perceived overweight on stress‐induced overeating via weight stigma concerns.”

Methods, Study 2

• The number of participants excluded for failing the attention checks changed from 97 to 104, bringing the sample size from 649 to 642.

• Chronbach α values for overeating tendencies and body dissatisfaction changed both to 0.92 from 0.91.

Results, Study 2

• Text should read, “In our first model (Table 3, Model 1), weight perception (perceived overweight relative to perceived normal weight) was a significant predictor of weight stigma concerns (B = 6.62; SE = 0.77; P < 0.001), and weight stigma concerns significantly predicted uncontrolled eating (B = 1.03; SE = 0.11; P < 0.001). There was a significant indirect effect of perceived weight on uncontrolled eating via weight stigma concerns (bootstrap estimate = 6.80; SE = 1.07; 95% CI: 4.85 to 9.06), and weight stigma concerns explained 60.3% of the variance in the relationship between perceived overweight and uncontrolled eating. In the fully adjusted model (Table 3, Model 2), perceived overweight, relative to perceived normal weight, had a significant indirect effect on uncontrolled eating via weight stigma concerns (bootstrap estimate = 1.49; SE = 0.55; 95% CI: 0.65 to 2.84), and weight stigma concerns explained 47.1% of the variance in the relationship between perceived overweight and uncontrolled eating. For gender, the index of moderated mediation was nonsignificant (bootstrap estimate = −1.07; SE = 0.84; 95% CI: −2.94 to 0.40), suggesting that gender did not moderate the indirect effect of perceived overweight on uncontrolled eating via weight stigma concerns. For weight perception accuracy, the index of moderated mediation was not significant (bootstrap estimate = −0.01; SE = 1.19; 95% CI: −2.94 to 1.91), indicating that weight perception accuracy did not moderated the indirect effect of perceived overweight on uncontrolled eating via weight stigma concerns.”

Discussion

Text should read, “In both studies weight stigma concerns explained a substantial proportion of the cross‐sectional association between weight perception and overeating in both our main analyses (32‐60% of variance) and in analyses that accounted for a range of other related psychological variables, including previous experience of weight discrimination (23‐47% of variance).”

The following corrections were made in the online Supporting Information

• Page 1: Study 1 measures, description of the variables income and education.

• Page 2: perceived weight discrimination measure reporting five items instead of six.

• Page 5: Correlation coefficients in Table S1 updated.

• Page 6: Study 2 measures: section added to describe subtle differences for Study 2 measures also used in Study 1.

• Page 8: Correlation coefficients in Table S2 updated.

The authors apologize for these errors and are now in the process of implementing additional policies in their research group that will prevent future error.

Supporting information