. Author manuscript; available in PMC: 2013 Jan 1.

Published in final edited form as: Acta Psychol (Amst). 2011 Oct 26;139(1):19–37. doi: 10.1016/j.actpsy.2011.09.014

Table 4.

Summary of models predicting log-odds accuracy based on various predictor sets.

	Model name	BIC	R²	Adjusted R²	RSE	F Statistic
1.	Full model	4162	.9823	.9616	.253	F(699,600) = 47.5
2.	Intercept + BIC-selected bias, similarity, and general perceivability	592	.666	.648	.259	F(66,1234) = 37.8
3.*	Intercept + BIC-selected bias, similarity, general & specific perceivability	572	.692	.672	.2498	F(77,1222) = 35.7
4.	Full bias + similarity (Biased choice rule)	4162	.845	.665	.253	F(699,600) = 4.68
5.	Bias + Similarity (BIC-Selected)	1000	.65	.617	.269	F(114,1185) = 19.43
6.	Bias + Perceivability (BIC-selected)	734	.577	.562	.289	F(42,1257) = 40.8
7.	Similarity + Perceivability (BIC-selected)	715	.597	.581	.283	F(48,1251) = 40.8
8.	Intercept-only model	1549	n/a	n/a	.4367	t(1299) = 100

Note: RSE = residual standard error (error sum of squares divided by the residual degrees of freedom). General similarity refers to a single set of similarity parameters fit across experiments. Specific similarity refers to using similarity parameters that can account for each experiment individually. Model 3, indicated with a *, indicates our preferred best model.