Tippett Plot
Definition
A calibration diagnostic plot that shows the cumulative distributions of LR values calculated for known same-source pairs and known different-source pairs. Well-calibrated systems show the two curves diverging strongly; overlapping curves indicate poor discrimination.
- Type
- Calibration diagnostic plot for likelihood ratio systems
- Curve 1
- LR distribution for known same-source pairs
- Curve 2
- LR distribution for known different-source pairs
- Well-calibrated result
- Curves diverge strongly rather than overlap
Common questions
What does it mean if the two curves on a Tippett plot overlap heavily?+
Heavy overlap means the system's likelihood ratios do not reliably separate true same-source comparisons from true different-source comparisons, so a given LR value from casework carries less discriminating power than a wider-separated pair of curves would give.
Who uses a Tippett plot and when?+
It is used during validation of a likelihood-ratio method, such as a geological or trace-evidence comparison system, before that method is relied on in casework, to demonstrate the system's real-world discrimination performance rather than just its theoretical design.
Related terms
- Between-Source Variation
- The variation observed among samples from different locations in the reference population. Used to estimate the probability of the observed similarity under...
- Denominator Hypothesis
- The defence hypothesis under which the LR denominator is calculated. Typically: the questioned sample comes from some other location within the relevant...
- ENFSI Verbal Scale
- A standardised table published by the European Network of Forensic Science Institutes that maps ranges of LR values to verbal expressions of...
- Numerator Hypothesis
- The prosecution hypothesis under which the LR numerator is calculated. In a source comparison it is typically: the questioned sample and the...
- Within-Source Variation
- The natural variation observed among multiple samples collected from a single source location. Used to estimate the probability of the observed difference...