Interpretation of Crime Scene Evidence Exam 3

strength of support (likelihood ratio)

Suppose a footwear examiner makes this statement in court: "In my opinion the correspondence between the footwear mark at the crime scene and the shoe of the accused is 4.5 times more likely when the shoe has made the mark [Proposition 1], than when the shoe has not made the mark [Proposition 2]". What kind of statement is this? 

Posterior, because the statement talks directly about the probability of the propositions themselves rather than the probability of the observations given the two propositions.

Consider this alternative phrasing: "In my opinion the correspondence between the footwear mark at the crime scene and the shoe of the accused means that the suspect's shoe [Proposition 1] is 4.5 times more likely to have made this impression than some other shoe [Proposition 2]". What kind of statement is this?

strong

Consider Table 1 from “The Expression and Interpretation of Uncertain Forensic Science Evidence: Verbal Equivalence, Evidence Strength, and the Weak Evidence Effect” paper. Suppose you have a likelihood ratio of 1500. What verbal expression (support) statement corresponds to this number? 

• Likelihood ratios are a measure of the discriminating ability of the evidence for two propositions, with is what the forensic scientist should be concerned with.

• Posterior beliefs are the job of the jury or other factfinder.

• You can make the prosecutor's fallacy where you equate the probability of evidence given innocence with the probability of innocence given evidence and ignores the base rates.

Why is it important to keep likelihoods and posterior beliefs separate if you are a forensic scientist testifying in court?

People often have a difficult time with numbers, especially extremely large or small ones.

Why are verbal expressions thought to be helpful?

The finding that weak inculpatory evidence makes jurors LESS likely to convict.

"weak evidence effect"

• Evidential strength (low, moderate, high)

• Presentation method (verbal vs numerical)

In the experiment “The Expression and Interpretation of Uncertain Forensic Science Evidence: Verbal Equivalence, Evidence Strength, and the Weak Evidence Effect”, which factors did they manipulate?

The difference between the amoung of guiltly vs innocent the defendant seemed after the evidence was presented, relative to how much guilty vs innocent they seemed before the evidence was presented.

Consider Figure 1 in “The Expression and Interpretation of Uncertain Forensic Science Evidence: Verbal Equivalence, Evidence Strength, and the Weak Evidence Effect”. What does the vertical axis represent?

The weak evidence effect

In Figure 1 of “The Expression and Interpretation of Uncertain Forensic Science Evidence: Verbal Equivalence, Evidence Strength, and the Weak Evidence Effect”, the first black bar goes negative? What is this consistent with?

Subjects undervalued the expert's testimony in all conditions

Implicit likelihood ratios divide the posterior belief odds (how much more guilty than innocent you think the defendant is AFTER you hear the evidence) by the prior belief odds (how much more guilty than innocent you think the defendant is BEFORE you hear the evidence). Recall that the Bayesian equation is Posterior Beliefs = LR * Prior Beliefs. Rearranging terms gives you the implicit likelihood ratio. What is the major finding from Figure 2 of “The Expression and Interpretation of Uncertain Forensic Science Evidence: Verbal Equivalence, Evidence Strength, and the Weak Evidence Effect”?

Subjects tended to change their beliefs less than intended by the testimony

(In “The Expression and Interpretation of Uncertain Forensic Science Evidence: Verbal Equivalence, Evidence Strength, and the Weak Evidence Effect”) In general, what was the degree of belief change in subjects relative to the intended strength of support in the testimony of the expert?

• People tend to undervalue forensic evidence when presented in strength-of-evidence form like likelihood ratios or verbal equivalencies.

• The weak evidence effect seems real and causes subjects to undervalue weak evidence to the point that they go in the wrong direction (i.e. inculpatory evidence is treated as exculpatory).

• There is no clear advantage of numerical or verbal equivalency forms of testimony in terms of having laypeople update their beliefs appropriately.

What are general issues that we need to think about from the “The Expression and Interpretation of Uncertain Forensic Science Evidence: Verbal Equivalence, Evidence Strength, and the Weak Evidence Effect” paper?

A specific sequence of bases at a particular locus.

Based on “Forensic DNA Analysis” What is analyzed in DNA evidence?

Versions of the DNA at an STR locus with different repeat lengths.

Based on “Forensic DNA Analysis” What are alleles?

Mutations create different number of repeats of STRs at a locus, and each one is an allele.

Based on “Forensic DNA Analysis” How is DNA used forensically?

• They help separate male from female DNA.

• They help find related males.

• They are inherited by sons from their fathers with very little mutation.

Based on “Forensic DNA Analysis” Which below is true for YSTR loci?

• The suspect is the source of the material.

• The match is a false positive due to contamination or some other kind of error.

• The material came from a second person who has an identical DNA profile to that of the suspect.

Based on “Forensic DNA Analysis” If a DNA analyst declares a "match" between two samples, which of the following could be true???

• The suspect touched another person who touched the surface.

• The suspect touched the surface.

• The suspect touched an object that was touched by a second person who then touched the crime scene surface.

Based on “Forensic DNA Analysis” If a DNA analyst declares a "match" between a crime scene sample and a suspect, which of the following could be true? 

prosecutor’s fallacy

if the probability of someone else is so low, then it must be the defendant

mean value fallacy

when someone tries to argue that the probability of a random match is smaller than the population and therefore it must be the defendant’s DNA

LR= 1/RMP

for single source DNA, what is the relation between the random match probability (RMP) and the likelihood ratio?

it helps determine whether the profile is missing data (or has incomplete data) due to amplification of low levels of material

What is the benefit of a stochastic threshold along with a control dot?

• mixture ratios

• a stochastic threshold

• an assigned number of contributors

• tow mutually exclusive propositions

• peak height

Which of the following information is used to interpret a mixture operation?

• the results are so clean with a single source sample that they either exclude or include a profile

• they rely on rules to handle stutter, peak height ratios, analytic thresholds

• they make statements directly about the probability of a proposition (included or excluded)

What is true with respect to “binary” interpretation?

posterior conclusion

makes a statement directly about the probability of a proposition

Probabilistic genotyping software (PGS)

uses biological modeling and statistical theory to calculate LRs for the DNA profile and the propositions of interest

• PGS can deal with mixtures with 3 or more contributors

• PGS can deal with low level samples

• PGS calculates a likelihood ratio, which computes the probability of the observations given 2 propositions

How does the PGS approach differ from the binary approach?

it should not replace trained DNA analysts

What did the authors and Gill et al. agree on with respect to PGS?

Different software produced variation in the results and their interpretation

What was the primary result of the MIX13 study, which was a black box study with mixtures that was designed to assess how good DNA analysts and PGS approaches functioned?

• The end user needs to understand the underlying mathematics and assumptions.

• DNA analysts will need to be trained in statistics and modeling in order to understand how PGS works.

• Without knowing how the software works to identify a suspect, the defendant runs the risk of being "convicted by code".

• The DNA analyst needs to be able to explain the basic functions of the software to a jury.

What is the problem with viewing PGS as a "black box"?

Subsource doesn’t care about the carrier fluid of the DNA

In DNA, what is the difference between source level and sub-source level propositions?

Hair color

In class we used a metaphor for DNA loci. What is the analog of a locus?

An STR will drop one repeat during amplification

In DNA, why does stutter occur?

The short tandem repeats at different loci

What parts of the DNA are used for forensics?

Knowing the alleles at one locus tells us nothing about the alleles at a different locus.

The loci at which we measure the STR values are chosen to be independent. What does this mean?