Fig. 1

Prevalence differences of class membership due to covariates; results of multinomial logistic regression from Minnesota National Guard Veterans who participated in the mailed survey conducted from October 2015 to September 2016. Interpreted as the difference in risk (probability) of membership in this latent class between levels of the covariate where a negative value means lower probability of membership in this class compared to probability at the reference level of the covariate; other covariates are standardized to their distribution among all responder. All covariates are dichotomous except for absorption, which is continuous. For absorption, a one-unit change represents a 2-point change on the absorption scale, equivalently a 7-point change on the T-score