Skip to main content

Table 1 Log-likelihoods and AICCs for the rank-ordered logistic regression models predicting participants’ rankings for each criterion set from their BWS or Likert scores. Higher log-likelihoods, and lower values of Akaike’s bias-corrected information criterion (AICC) (i.e. closer to zero for both measures) indicate better model fit. Models using best-worst scaling (BWS) scores achieved better fit for each of the six criterion sets. This improvement is reflected by the difference between AICC values for the two models (i), which are > 10 for all criterion sets, indicating a substantial improvement in model fit (Symonds & Moussalli, 2011)

From: Best-worst scaling improves measurement of first impressions

Criterion set BWS Likert i
Log-likelihood AICC Log-likelihood AICC
1 − 528.74 1059.49 − 562.19 1126.39 66.90
2 − 317.47 636.95 − 366.44 734.89 97.94
3 − 509.45 1020.91 − 563.55 1129.11 108.20
4 − 400.60 803.21 − 478.17 958.35 155.14
5 − 494.29 990.59 − 548.38 1098.77 108.18
6 − 489.88 981.77 − 561.11 1124.23 142.46