Patricia Carlin1, Paul Barrett2, and Gisli Gudjonsson3
1&3 Institute of Psychiatry 2Psytech International Ltd. 2University of AucklandDept. of Psychology The Grange Faculty of BusinessDe Crespigny Park Church Road, Pulloxhill Commerce C BuildingDenmark Hill Bedfordshire Symonds StreetLondon SE5 8AF, UK MK45 5HE, UK Auckland, NZ
2University of CanterburyDept. of PsychologyPrivate Bag 4800Christchurch, NZ
Patricia Carlin1, Paul Barrett2, and Gisli Gudjonsson3
1&3 Institute of Psychiatry 2Psytech International Ltd. 2University of AucklandDept. of Psychology The Grange Faculty of BusinessDe Crespigny Park Church Road, Pulloxhill Commerce C BuildingDenmark Hill Bedfordshire Symonds StreetLondon SE5 8AF, UK MK45 5HE, UK Auckland, NZ
2University of CanterburyDept. of PsychologyPrivate Bag 4800Christchurch, NZ
NZ Psych Conference – Aug/Sept 2003
Download from: www.pbarrett.net
The Psychopathy Checklist Revised (PCL-R: Hare, 1991, 2003) provides three scores to be used in the assessment and diagnosis of psychopathy as well for actuarial-predictive risk applications:
NZ Psych Conference – Aug/Sept 2003
a Factor 1 “interpersonal variables” scorea Factor 2 “social deviance” scorea Total score formed from the sum of all
20 items comprising the PCL-R.
NZ Psych Conference – Aug/Sept 2003
It consists of 20 cognitive-behavioural constructs or items, each assigned a 0, 1, 20, 1, 2rating indicating the degree to which an individual matches the description of the item in the test manual. A 0 rating indicates that the item does not apply to the individual. A rating of 1 indicates a partial application, with 2 indicating that an item definitely applies to an individual.
NZ Psych Conference – Aug/Sept 2003
The sum score across all 20 items yields the total PCL_R score (0-40). Two other scores are normally computed, the Factor 1 and 2 scores (F1 & F2) which are the summed scores respectively for the particular items …
NZ Psych Conference – Aug/Sept 2003
Failure to Accept Responsibility for Own Actions
16
Callous/Lack of Empathy8
Shallow Affect7
Lack of Remorse or Guilt6
Conning/Manipulative5
Pathological Lying4
Grandiose Sense of Self Worth2
Glibness/Superficial Charm1
Factor 1: Interpersonal VariablesPCL-R itemreference ID
*For confidentiality reasons, these items do not appear in the web version of the paper
NZ Psych Conference – Aug/Sept 2003
Criminal Versatility *new in 2003 manual20
Revocation of Conditional Release19
Juvenile Delinquency18
Irresponsibility15
Impulsivity14
Lack of Realistic, Long-Term Goals13
Early Behavioural Problems12
Poor Behavioural Controls10
Parasitic Lifestyle9
Need for Stimulation/Proneness to Boredom
3
Factor 2: Social DeviancePCL-R itemreference ID
NZ Psych Conference – Aug/Sept 2003
It is proposed by Hare that the factors represent two distinct but related aspects of psychopathy.
Hare, in the 1991 manual, p.38 states “The “The items that make up the two PCLitems that make up the two PCL--R factors, which R factors, which on average correlate about on average correlate about 0.50.5 with one another”with one another”.
In the 2003 manual, p. 77, “For both the PCL “For both the PCL and PCLand PCL--R the correlation between factors 1 and R the correlation between factors 1 and 2 was about 2 was about 0.50.5””.
NZ Psych Conference – Aug/Sept 2003
Facet 1: Interpersonal1. Glibness/Superficial Charm2. Grandiose Sense of Self-Worth4. Pathological Lying5. Conning/Manipulative
Facet 2: Affective6. Lack of Remorse or Guilt7. Shallow Affect8. Callous/Lack of Empathy9. Failure to Accept Responsibility forOwn Actions
Facet 3: Lifestyle3. Need for Stimulation/Proness to Boredom9. Parasitic Lifestyle13. Lack of Realistic/Long-Term Goals14. Impulsivity15. Irresponsibility
Facet 4: Antisocial10. Poor Behavioural Controls12. Early Behavioral Problems18. Juvenile Delinquency19. Revocation of Conditional Release20. Criminal Versatility
Factor 1 Factor 20.699
0.809
0.691
0.752
0.710
The 2003 2nd Edition model
NZ Psych Conference – Aug/Sept 2003
Given an expected observed correlation between the two factor scores, to what extent might disparities between the two scores exceed “permissible” values?
Legal Context:defence –v- prosecution psychologists rating the same offender – but with markedly different ratings on Factor 1 – leading to quite different total scores.
NZ Psych Conference – Aug/Sept 2003
Evaluate the expected distributions of scores on each factor, conditionalconditional upon the scores of the respective factor using a sample of10,000 simulated F1 and F2 scores which:
possess a correlation of 0.5are drawn from perfectly normally
distributed distributions (bivariate normality as required for a Pearson correlation coefficient)
possess means and SD’s as per 2003 test manual.
NZ Psych Conference – Aug/Sept 2003
Evaluate the expected distributions of scores on each factor, conditionalconditional upon the scores of the respective factor using two samples of actual data:
Forensic psychiatric patientsForensic psychiatric patients (N=217)Prisoners/OffendersPrisoners/Offenders (N=1358)
NZ Psych Conference – Aug/Sept 2003
Forensic psychiatric patientsForensic psychiatric patients (N=217) from two of the UK’s High Security Forensic Psychiatric Hospitals, Ashworthand The State Hospital) and Arnold Lodge (a high security clinic).
Prisoners/OffendersPrisoners/Offenders (N=1358) from several UK prison institutions (some data donated by David Cooke).
NZ Psych Conference – Aug/Sept 2003
The study participants were all male adultsmale adultswho had been assessed using a PCL-R within a personal interview (in addition to collateral file information) - conducted by a formally trained practitioner.
NZ Psych Conference – Aug/Sept 2003
Where missing ratings were encountered, the Factor and Total scores were prorated as per manual instructions.
All scores were created via computer from raw item data, using a custom scoring routine which handled prorating and the computation of F1, F2, and Total scores.
NZ Psych Conference – Aug/Sept 2003
Step 1: Construct two variables, Norm_1 and Norm_2, sampling 10,000 observations from a random normal population distribution with mean of 0.0 and standard deviation of 1.0.
Step 2: Construct a new variable, New_Norm_2, from the two variables in step 1 according to the formula …
NZ Psych Conference – Aug/Sept 2003
( )2_ _2 _1 + _2
1
rNew Norm Norm Normr
⎛ ⎞⎜ ⎟= ⋅⎜ ⎟⎜ ⎟−⎝ ⎠
where:
Norm_1 and Norm_2 are random normally distributed sets of continuous-valued “scores”
r is the desired correlation between Norm_1 & New_Norm_2 (0.54*) *allowing for integer mapping & scaling constraint
attenuation which will bring the correlation back to 0.5
New_Norm_2 is the transformed Norm_2 variable
NZ Psych Conference – Aug/Sept 2003
Step 3: Convert1 the real-valued variables into integer “factor” scores, designed to possess the same mean and standard deviation as provided in Table 4.7 of the PCL-R manual (Hare, 2003). This was achieved separately for Offenders and Patients as their means and SDs are quite different. 1(note:note: this required two kinds of rescaling adjustments, so as to allow for the constrained integer measurement range of the PCL-R).
191020814516
191020229015
191020236314
18920356513
18920172812
18820284111
17820297210
17720011339
16620010318
1652009407
1552008356
1442006705
1432005194
1332004113
1321902922
1221601761
111160890
90th
Percentile Score on Factor 2
10th
Percentile Score on Factor 2
Maximum Observed Factor 2
score
Minimum Observed Factor 2
score
Number of cases scoring at this level
Factor 1 Score
Offender Data
15716320020
15716226119
14716235418
14716442417
14716060016
14616269315
13516073414
13516183113
13516088412
12416085811
13416083510
1231607989
1131606108
1021505107
1021504296
911502915
911402494
911201873
811501342
70120641
7090540
90th Percentile Score on Factor 1
10th Percentile Score on Factor 1
Maximum Observed
Factor 1 score
Minimum Observed
Factor 1 score
Number of cases scoring at this level
Factor 2 Score
NZ Psych Conference – Aug/Sept 2003
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Factor 1 Score
0
100
200
300
400
500
600
700
800
900
1000
1100
1200
No.
of o
bser
vatio
ns
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Factor 2 Score
0
100
200
300
400
500
600
700
800
900
1000
No.
of O
bser
vatio
ns
Simulated Data, N=10,000 casesWith F1 v F2 correlation of 0.50
NZ Psych Conference – Aug/Sept 2003
Step 1: Look at the data for patients and offenders – to see whether they can be sensibly combined into one sample …
NZ Psych Conference – Aug/Sept 2003
0.500.500.50Factor r
4018182016Maximum
00000Minimum
8.14.44.44.93.7SD
158895Median
15.78.08.08.95.3Mean
Total Old Factor 2
Factor 22003
Factor 1
OffendersN = 1358
Male
NZ Psych Conference – Aug/Sept 2003
0.310.310.28Factor r
3517171916Maximum
10000Minimum
6.93.83.84.34.0SD
201111117Median
19.79.99.910.87.4Mean
Total Score
Old Factor 2
Factor 2Factor 1
PatientsN = 217
Male
Conclusion: Patients and Offenders to be analysed separately
NZ Psych Conference – Aug/Sept 2003
Step 2: Use resampling analysis to provide “population” estimates of F1 and F2 conditional score distributions.
NZ Psych Conference – Aug/Sept 2003
Step 3: Construct 17 datasets of ~1000 randomized scores constructed in line with the observed proportions of each F2 score at each F1 score point (0-16 range -F1 scores).
Step 4: sample N (=total dataset) cases at random (uniform sampling) “with replacement”from F1 scores in the same dataset from which the F2 conditional score distributions have been calculated.
NZ Psych Conference – Aug/Sept 2003
Step 5: For each resampled F1 score, sample an F2 score at random from the proportionate conditional F2 score distribution associated with a particular F1 score magnitude (a secondary conditional resampling)
Step 6: Take 1000 complete resamples of the F1 dataset (1000 x 217 cases in respect of the patients, and 1000 x 1358 cases in respect of the offender data).
NZ Psych Conference – Aug/Sept 2003
Step 7: Compute the overall conditional distributions of F2 scores conditional upon each F1 score, and F1 scores conditional upon each F2 score, using all 1000 samples
18101810704116
1851951197215
1982061685114
1661742824813
1751823310512
1772014974811
1751805473810
173200680339
164201785508
1532011077167
1441801122276
1421701400025
1431801399664
1211801765133
1211701370552
1101501069391
90170892960
90th
Percentile Score on Factor 2
10th
Percentile Score on Factor 2
Maximum Observed Factor 2
score
Minimum Observed Factor 2
score
Number of cases
scoring at this level
Factor 1 Score
Offender Data Resample
147147709820
148158888019
1441632601518
1431603961517
1331605822716
1241414606615
1231607740914
1031617878313
921409279012
9113010376311
10216110627110
91120886529
91150780708
81150895187
90140872676
80150620855
70130750774
70110611873
70120758532
60110563281
100100390460
90th Percentile Score on Factor 1
10th Percentile Score on Factor 1
Maximum Observed
Factor 1 score
Minimum Observed
Factor 1 score
Number of cases scoring at this level
Factor 2 Score
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Factor 1 Score
0
100
200
300
400
500
600
700
800
900
1000
1100
1200
No.
of o
bser
vatio
ns
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Factor 2 Score
0
100
200
300
400
500
600
700
800
900
1000
No.
of O
bser
vatio
ns
Simulated Datasets – Normative (manual) Offender Specifications
UK offenders (N=1358) actual data
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Factor 1 Score - Offender Data
0
20
40
60
80
100
120
140
160
180
200
No.
of O
bser
vatio
ns
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Factor 2 Score - Offender Data
0
10
20
30
40
50
60
70
80
90
100
110
120
No.
of O
bser
vatio
ns
Top Related