240 likes | 393 Views
Measuring Subjective P robability in the Context of Medication R isks Niels Haase , Frank Renkewitz & Cornelia Betsch University of Erfurt (Germany). Measures of Subjective Probability. 7-Point Rating Scale. 11-Point Rating Scale. Visual Analog Scale.
E N D
Measuring Subjective Probability in the Context of Medication RisksNielsHaase, Frank Renkewitz &Cornelia BetschUniversity of Erfurt (Germany)
MeasuresofSubjectiveProbability 7-Point Rating Scale 11-Point Rating Scale Visual Analog Scale
MeasuresofSubjectiveProbability Frequency Format Percent Format
MeasuresofSubjectiveProbability • Differencesbetweenscales • Resolution • Verbal vs. numeric • Evaluative criteria • Usability • Test-retestreliability • Subjectiveconfidence in judgment • Behaviorprediction
Research Question & Design • How do thesescalesperformwhenobjectiveprobabilitieshavebeenlearned? • Sensitivity Individualcorrelation between objective and subjective probabilities • ContextDependency Differentiation betweenrangesofprobability
Hypotheses • Rating scales’ disadvantage due to low resolution and vague category quantifiers • Visual analog scale’s disadvantage due to difficulties in its use • H1: Numeric scales show higher sensitivity than the visual analog scale and the rating scales. • H2: Rating scales and the visual analog scale will be more context-dependent.
Independent Variables • 5 Scale Formats (between) • 2 Rangesofprobability (between): 2%, 5%, 10%, 20% vs. 42%, 45%, 50%, 60% • Mode ofpresentation (within): Sequential Graphical (control)
SequentialEncoding: Sensitivity Low Range ofProbabilities: 2%-20% High Range ofProbabilities: 42%-60%
SequentialEncoding: ContextDependency 7-point ratingscale Percentformat 20% 42% 20% 42%
Summary: SequentialEncoding • Allscaleswerehighlysensitive • Numericscalesweregenerallysuperior • Rating andvisual analog scaleswerecontextdependent • Differencesbetweenscalesarereducedwhenencodingismoreerror-prone
GraphicalPresentation Probabilityof 42%:
GraphicalEncoding: Sensitivity Low Range ofProbabilities: 2%-20% High Range ofProbabilities: 42%-60%
GraphicalEncoding: ContextDependency 7-point ratingscale Percentformat 20% 42% 20% 42%
Summary: GraphicalEncoding • Altogetherhighersensitivity • Differencesbetweenmeasuresremainthesame, numericformatssuperior • Reducedcontextdependency • Noeffectofrange, i.e. encodingdifficulty
Experiment 2: Effectofencodingerror • Encodingmoreerror-prone • 5 outcomeswithin 1 sequence • Shorter inter-stimulus time • 3 repeatedmeasurements • 2 Scaleformats (between) 7-point ratingscale vs. Percentformat • 2 Rangesofprobability (within): 10%, 20%, 30%, 40%, 50% vs. 50%, 60%, 70%, 80%, 90%
Experiment 2: Sensitivity Small Probabilities: 10%-50% Large Probabilities: 50%-90%
Experiment 2: ContextDependency 7-point ratingscale Percentformat Small Probabilities: 10%-50% Large Probabilities: 50%-90% 10% 20% 30% 40% 50% 50% 60% 70% 80% 90% ObjectiveProbabilities 10% 20% 30% 40% 50% 50% 60% 70% 80% 90% ObjectiveProbabilities 50% 50% 50% 50%
Conclusions & Implications • Different scale formats vary in sensitivity with numeric measures generally faring the best • Sensitivity also varies as a function of error in infomation at the time of judgment • Scale characteristics (e.g. resolution) become less important with increasing error • Context dependency increases with increasing error • In the realm of probabilities rating scales are not suitable for between-subjects research