Verbal Probabilities: Linear or Logistic? – A Regression Analysis Approach
Thomas Ostermann, Vogel Hannah, Appelbaum Sebastian · Studies in health technology and informatics · 2018
Verbal probability expressions are quite intuitive and used in a variety of clinical important issues like adverse drug causality appraisal. However, there is insufficient evidence of their numerical meaning and whether they have a linear or logistic relationship with them. We aimed at contributing to answering these questions by means of a comparative regression analysis based on a sample of N=683 participants between 10 and 82 years (mean age 20.33±11.77; median: 18 years) who were asked to numerically rate a given set of sixteen verbal probability phrases on a visual analogue scale. With respect to the explained variance, we found an R2for the linear model of 0.574 while R2for the logistic model was only 0.392 indicating a superiority of linear model compared to the logistic model. Although we were able to show that ranked verbal phrases are more likely to behave in a linear that in a logistic way other regression options like the double logistic model should be taken into consideration for further research.