Table 3:

Logistic Regression and Negative Binomial Regression Models Comparing Athletes and Non-athletes Across Disordered Eating Engagement and Episode Frequency

	Logistic Regression Model			Negative Binomial Regression Model*
Disordered Eating Behavior	Sample Size (n missing)	Odds Ratio [95% CI]	q-value	Sample Size Reporting at Least 1 Episode in Past 3 Months	Incidence Rate Ratio [95% CI]	q-value
Binge Eating	23,920 (0)	0.85 [0.78, 0.92]	<.001	17,903	0.91 [0.87, 0.95]	<.001
Vomiting	23,920 (0)	0.92 [0.84, 1.00]	.071	6,231	0.97 [0.88, 1.06]	0.545
Diuretics/Laxatives	23,907 (13)	0.97 [0.87, 1.07]	.545	4,053	0.91 [0.82, 1.02]	.140
Excessive Exercise	23,899 (21)	2.30 [2.13, 2.49]	<.001	11,596	1.40 [1.32, 1.47]	<.001
Fasting	23,805 (115)	0.78 [0.73, 0.85]	<.001	12,068	0.99 [0.94, 1.05]	.826

q-values and associated odds ratios or parameter estimates that are bolded are significant results after FDR corrections.

^*Negative binomial regression models only included the subset of the sample that reported at least 1 episode of the respective disordered eating behavior. Therefore, there were no missing data in each negative binomial regression model.

df=degrees of freedom, CI= Confidence Interval