Table 1. Exploration of the dimensionality of the attitude-toward-the-two-candidates items.

Model	G2	npar	rD1-D2	relD1	relD2	relD3
Notes. G2 stands for the model fit statistic (its deviance), which is a log-likelihood statistic multiplied by −2 (see Adams, Wilson, & Wang, 1997). Differences in deviances are generally assumed to have a χ2-distribution with the difference in the number of parameters (npar) as degrees of freedom. Thus, we can statistically compare model fits of, for example, the less restricted two-dimensional (2D) model with the simpler one-dimensional (1D) model. Note that superior fit shows in smaller deviances. rD1-D2 = correlation between Dimension 1 and Dimension 2; rel = reliability estimates for Dimensions 1, 2, or 3. The correlations are estimates that have been disattenuated for measurement error (for technical details, see Adams et al., 1998).
1D: pro-R-anti-M	90,311.86	45	–	.92	–	–
2D: pro-R & pro-M	83,032.13	47	−.44	.91	.88	–
3D: 2 D & Feeling	87,555.12	53	−.31	.87	.79	.64