For two-way contingency tables, Goodman (1979) considered the local row effects (R) model and Agresti (1984, p.125, p.147) considered the logit row effects (LR) model and the global row effects (GR) model. This paper proposes three measures which represent the degrees of departure from the R, LR, and GR models. The proposed measures are expressed as a minimum Euclid distance between (i) the local odds-ratios, (ii) the local- global odds-ratios, and (iii) the global odds-ratios, and the corresponding odds-ratios with the structure of R, LR, and GR, respectively. The measures would be useful for comparing the degrees of departure from each row effects model in several tables. Examples are given. Note that the proposed measures are extensions of Altham's (1970) measure of departure from the null association model, and Tomizawa's (1993a) and Tomizawa, Seo and Fujino's (1995) measures of departure from the local, logit, and global uniform association models.