Nicotine dependence studies are increasingly combining samples of participants to increase statistical power and make comparisons across groups of diverse age, race/ethnicity, and sex (Belsky et al., 2013; Bierut et al., 2007; Fagerstrom and Furberg, 2008; John et al., 2003). In studies that compare an underlying latent trait, like nicotine dependence, it is assumed that the instrument is measuring the trait on a consistent scale (i.e., it is invariant, measuring the trait similarly across groups) (Widaman and Reise, 1997). Measurement non-invariance is a type of measurement error that can bias study results toward or away from the null hypothesis, thereby leading to incorrect results in statistical comparisons and increasing the chances of both Type 1 and Type 2 errors. A non-invariant measure of nicotine dependence might incorrectly suggest that groups differ in their dependence levels (Schroeder and Moolchan, 2007) or the relation between dependence and other key variables (e.g., estimating the association between nicotine dependence and cessation, in which cessation is correlated with group membership like age (Johnson et al., 2008)). Measurement invariance may also obscure true associations, making them
