Scribd
Upload
420 views

Hare Psychopathy Checklist-Revised

Uploaded by

Phureephak Jinopeng
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
420 views

Hare Psychopathy Checklist-Revised

Uploaded by

Phureephak Jinopeng
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
You are on page 1/ 11
Evaluating the Screening Version of the Hare Psychopathy Checklist—Revised (PCL:SV): An Item Response Theory Analysis David J. Cooke and Christine Michie Glasgow Caledonian University Stephen D. Hart Simon Fraser University Robert D. Hare University of British Columbia ‘The Serconing Version ofthe Psychopathy Checklist Revised (PCL:SV; 8. D- Har, D.N. Cox, & RD. Hare, 1995) was developed to complement the Psychopathy Chookist~ Revised (PCL-R; R. D. Hare, 1991), and for use outside forensic setings. The PCL:SV takes less time to administer and requires less collateral information than the PCL-R. An item respoase theory approach was adopted to determine similarities inthe structural properties ofthe 2 instruments and whether the PCL:SV could be regarded as short form of the PCL-R. Eight of the 12 items in the PCL:SV were strongly parallel to thie equivalent PCL-R items. OF the 4 items PCL:SV items which differed from their equivalent PCL-F items, all 4 were found tobe equal or superioe to their equivalent PCL-R items in terms of discrimination, ‘Tre analyses confirmed previous results thatthe interpersonal and affective features of psychopathy have higher thresholds than do the impulsive and antisocial behavioral features individuals have t Be at 2 higher lev! of the psychopathic trait before the interpersonal and affective features become evident. The PCL:SV isan effective short form ofthe PCL-R. ‘The Hare Psychopathy Checklist—Revised (PCL-R; Hare, 1991) is a rating scale designed to measure traits of psychopathic personality disorder in forensic populations. It comprises 20 items, each reflecting a different symptom or characteristic of psychop- athy (See Table 1). Items are defined in detail in the PCL-R ‘manual and are rated on a 3-point scale (0 = item doesn’t apply, 1 = item applies somewhat, 2= item definitely applies). The items are summed t0 yield total scores, ranging from 0 to 40, that reflect the degree to which an individual resembles the prototypical psychopath. A cutoff score of 30 or greater is used to diagnose psychopathy. The items also can be summed to yield scores on two moderately correlated factors (Cooke, 1995; Hare et al., 1990; Harpur, Hakstian, & Hare, 1988; Harpur, Hare, & Hakstian, 1989) Factor | reflects the affective and interpersonal features of psy- chopathy and has been labeled the Selfish, Callous, and Remorse- less Use of Others; Factor 2 reflects the social deviance features of psychopathy and has been labeled, Chronically Unstable and An tisocil lifestyle (Hare etal, 1990). ‘The PCL-R was developed originally for use in experimental psychopathology, and it has yielded a large and impressive body of ‘David J. Cooke and Christine Michie, Deparment of Psychology, Gla sow Caledonian University; Stephen D. Hart, Department of Psychology Simon Fraser University, Burnaby, British Columbia, Canade; Robert D. Hare, University of British Columbia, Vancouver, Briish Columbia, Canada, ‘Correspondence conceming this article should be addressed to Davi J. ‘Cooke, The Douglas Inch Cente, 2 Woodside Terrace, Glasgow, G3 TUY, United Kingdom. Electonie mail may be sent © djooke@rgardens research (see Cooke, Forth, & Hare, 1998 for a review). However, more recently, the PCL-R has received considerable attention from forensic researchers because of is predictive validity with respect to criminal behavior, and in particular violent crime (for recent reviews, see Hart & Hare, 1996, 1997; Salekin, Rogers, & Sewell, 1996). For example, the PCL-R is related to failure on conditional release (eg., Hart, Kropp, & Hare, 1988; Hemphill, Hare, & Wong, 1998; Serin, Peters, & Barbaree, 1990), violent recidivism (Haris, Rice, & Quinsey, 1993; Quinsey, Rice, & Harris, 1995; Serin, 1996; Serin & Amos, 1995), and poor treat ment response (Ogio, Wong, & Greenwood, 1990; Rice, Haris, & Cormier, 1992). Reviewers have commented favorably on the PCL-R's psychometric properties and its criterion- and construct- related validities (e.g, Fulero, 1995; Rogers, 1995; Stone, 1995). One problem with the PCL-R is that administration requires access to detailed interview and case history information, inelud- ing a criminal record, This means that the PCL-R is costly to administer in terms of time and effort, and it also may be inap propriate for use outside of forensic settings where criminal records often are unavailable or irelevant (thus rendering several items unscoreable). These consideration led to the development of the Screening Version ofthe PCL-R (PCL:SV: Hart, Cox, & Hare, 1995). The PCL:SV is 12-item rating scale based directly on the PCL-R. The item descriptions in the PCL:SV manual are very brief relative to those for the PCL-R and require less detailed information to score. Most of the PCL:SV items were derived 80) with the PCL-R in several samples, even when the scales were administered on sep- state occasions by independent raters. Evaluating the PCL-R Using Item Response Theory Most evaluations of the PCL-R’s psychometric or structural properties have been based on classical test theory. In genera, {these evaluations have concluded thatthe PCL-R seems appropri ate for use with a wide range of forensic populations, including ‘mentally disordered offenders, a variety of different ethnic and linguistic minorities in North America and offenders in several European nations (e.., af Klinteberg, Humble, & Schalling, 1992; Cooke, 1995, 1998; Coté, 1990; Forth, Hart, & Hare, 1990: Haa- pasalo & Pulkkinen, 1992; Hare et al. 1990; Hart & Hare, 1989: Kosson, Smith, & Newman, 1990; Raine, 1985; Wong, 1984), However, Cooke and Michie (1997, in press) have argued that item, response theory has several advantages over classical test theory a8 a framework for evaluating the PCL-R, and particularly in eval- ating its suitability for rosscultural research. Item response the- ‘ry approaches to the analysis of test items and test functioning are , parameters ae endorsed positively only when 0s relatively large, and increases in b, result in the shifting of the ICC to the right, The b, paramcters are the item response theory equivalent of the proportion of comect responses, op in clasical test theory CTT. Data Analyses Our evaluation ofthe PCL:SV required two general steps. First, we constructed separate item response theory models for the PCL:SV and PCL-R, This required demonstrating tha the PCL:SV and PCL-R were unidimensional sales, as our selected mode! asumes that @isunidimen- sini this is variously refered to asthe assumption of homogene, local Independence, or conditional independence’ Aso inthis sep, we tested the generality of the item response theory models across the gender of subjects In the second step. the one of primary interest in this article, we ‘compared the item response theory models forthe PCLSV and PCL-R. At the item evel, we examined the similarity of the ICCs for corresponding PCL:SV and PCL-R items (or item pais). tthe level of composite scores, we examined the relationship berween the latent (05) measored by the PCL:SY and PCL-R items. Next, we compared the test and item informa tion fonctions for the PCL:SV and PCL-R. Finaly, we determined the agnostic efficiency of the PCL:SV with espect to the PCL-R. Allitem response theory analyses were conducted using Mullog (This: sen, 1991), 0 well-rarded and widely used computer program (©. Nunnally & Bernstein, 1994) The model iting procedure i a estimation ‘minimalizaon (EM) algorithm developed by Bock and Aitken (1981). Results Unidimensionality There is a lack of consensus concerning how best to assess dimensionality Hate, 1985). Traditionally, it has been evaluated using Cronbach's a, mean r,, othe relative proportion of variance accounted for by the first and second unrotated principal compo- nents. Using these criteria, we found the PCL:SV appeared to be ‘unidimensional in the current sample of $86 participants: Cron- bach’s a was 96, mean 7, was 66, and the ratio between the fist and second-unrotated principal components was about 5:1. Hare (1991) argued that the PCL-R also is essentially homogeneous, with a second-order factor being underpinned by two correlated first-order factors. Analysis of the PCL-R ratings in this data set supported Hare's findings: Cronbach's a was .93, mean r,, was 60, and the ratio of the variance accounted for by the first and second unrotated principal components was about 6:1. In light of these results, we concluded that the PCL:SV and PCL-R were sufficiently unidimensional to permit their evaluation using item response theory methods. Confirmatory factor analysis has dem onstrated that both instruments {it a hierarchical model with & highly saturated superordinate factor overpinning first-order fac- tors (Cooke & Michie, 1998) Generality of Item Response Theory Models Because of limited sample size, it was impossible o fit item response theory models tothe data from cach individual subsample in the data set, Therefore, we were unable to investigate whether the item response theory parameters were the same actoss settings. However, an attempt was made to develop separate PCL:SV ‘models for men and women. By comparing these models, we ‘would be able to determine whether there was evidence of differ- ential item functioning (DIF). DIF would be suggested by the presence of significant differences in the value of item parameters across gender. Inthe PCL:SV normative data set, complete ratings were available for 405 men and 181 women. Using Multilog, we fitted a three category version of Samejima’s (1969) graded model to these data by the method of maximum likelihood. The run for ‘men was successful; however, the run for women failed to con: verge because « matrix became singular. Ths singularity was most, probably a consequence of the limited range of 0 in the female sample, and the large number of parameters being estimated (36 item parameters) and the small sample size (n = 181). To combat, this problem, we compared the item parameters for men and women in batches of six items at a time: the first six items, the riddle six items, and the last six items. We found significant differences (p < .05) in te first of these three comparisons. On inspection, these differences appeared to be due to a large differ ence in the b, parameter for Item 5, Lacks Empathy. As the observed differences are due to only one item parameter and may well have been the consequence of multiple-sgnificance testing, we made the assumption—on the principle of parsimony—that, there was no DIF across gender in this sample. However, the sample size was relatively small; its possible that future analyses based on larger samples will find DIF. It was not possible 10 examine DIF across gender for the PCL-R, due to small sample Comparison of PCL:SV and PCL-R Items ‘Maultlog was used to construct a graded model that contained the 12 PCL:SV items and the 20 PCL-R items. Estimating the parameters ofall 32 items simultaneously ensures that we have & common metric, and thus, tis valid to compare item parameters or estimates of trait. The fitted parameters for this unconstrained model are presented in Table 3. To facilitate comparisons, corr sponding PCL:SV and PCL-R items are aligned Is apparent from Table 3 that the parameters of corresponding items are similar in many instances. The method of generalized likelihood ratio testing (GLRT) was used to investigate the degree of similarity. According to the theory of GLRT, under certain conditions, imposing constraints on such a model, for example, constraining certain parameters to be equal, leads to an increase in the statistic G? = ~2 log likelihood. The increase is distributed as 4 statistic with degrees of freedom equal to the number of constraints imposed. A series of nested analyses was caried out. At each step in the "This is nota song assumption, however, a8 IRT models ae relatively robust to departures from unidmensioaality. Indeed, seme IRT methods explicitly take into account and analyze multidimensional (2. Sti herp & Thissen, 1996; Yen, 1993). Probabilty(response) Figure 1 IRT ANALYSIS OF THE PCL:SV PCL:SV Doesn't Accept Responsibility Trait Example of item response theory curves. PCL:SV Screening Version of the Paychopsthy Checklist Revised nested analyses, the parameters of an additional pair of corre. sponding items were constrained to be identical. The resulting increase in G? over that for the unconstrained model was tested, ‘and if not significant, then the item pair remained constrained in the next analysis. The analyses were carried out in four stages. At the first stage, item pairs in which a PCL:SV item corresponded with only one PCL-R item were constrained. At the second stage, were constrained tobe identical. At the third stage, the two PCL-R pairs that did not cause a significant increase in G were con strained to have identical parameters tothe corresponding PCL:SV item, that is, all three items were constrained to have equal pa rameters. At the fourth stage, we investigated the four PCL-R pairs that did not have equivalent parameters. In these cases, the parameters of the PCL:SV item were constrained to be equal 10 the pairs of PCL-R items that corresponded with one PCL:SV item item parameters of each ofthe PCL-R items in turn. The sequence Table 3 Comparison of Hare Psychopathy Checklist: Screening Version (PCL-SV) and Revised (PCL-R) Items: Full Model PeLsv PCL-R em @ by by a bs by Superficial 1S O18 Superficial 16 021s Grandoe 16-0213. Grandiose 030A Devetfal 17-04 13 Conning/manipulstive te 0s 12 Pathological ying 1s 0016 Lacks remorse 32-05 06 Lack of remorse 24-0904 Lacks empathy 29 05 1 Callousck of empathy 35-0308 Shallow affect 21-08 08 Doesn't accept responsibilty «24-706 lle to accept responsiblity for own actions. © «LT. -09 OS Impulsive 20-1202 mpasivity 24 1003. Need for stimulation 20 i Oa Poor behavior contol 16-08 OS _—_Poor behavior controls is 06 (06 Lacks goals 16 12 08 Lack of goals Is -10 06 Paras lifestyle Z1 -06 09 Irresponsible 23-1004 responsibility 26 1008 ‘Aidolescent amsocial behavior d= 08 (O8——_—Juvenile delinquency Io eearoomen 4 Early behavior problems rs or ta Adult antisocial behavior 25-08 -03_—_—Revocation of conditional release re -04 08 Criminal versity 2 02 os ‘Note. «isthe slope ofthe logit curve at the point of infleton; by and b, ae the thresholds for Py and P.,respestively 8 of the analyses and the associated statistics are presented in Ta- ble 4. The item parameters of the resulting model with the con- straints that did not result i significant increases in G? are given in Table 5. The mode! fits the data well, predicting the observed patter of responses for each item within 1%. In Table 5, corre- sponding items that have been constrained to have identical pa- rameters aze shown in boldface type. Tn sum, the nested analyses suggested that 8 of 12 PCL:SV items (tems 1, 2, 3,6, 7, 8, 9, and 10) had parameters identical to the PCL-R items from which they were derived, Thus, these items can be considered strongly parallel to their PCL-R coun- terpatts, The remaining 4 PCL:SV items (Items 4,5, 11, and 12) had parameters that were significantly different from the cor- responding PCL-R items. We decided to focus on these non parallel items to determine whether their inclusion in the PCL:SV was problematic. GLRT indicated only tha item parameters were different; i did not reveal how the items differed. Inspection of the parameters for the 4 nonparalle items (see Table 5) indicated that PCL:SV Item 4 (lacks remorse) was more discriminating and had higher thresholds than its corresponding PCL-R item, PCL:SV Item 5 (lacks empa- thy) was more diseriminating than one member of the correspond- Jing PCL-R item pair, but less discriminating than the other Finally, PCL:SV Items 11 (adolescent antisocial behavior) and 12 (adult antisocial behavior) were equally as discriminating as the members of their corresponding PCL-R item pairs and had lower thresholds. These findings indicate that the four nonparallet PCL:SV items are equal or superior to their PCL-R counterparts with respect to discrimination, and 3 of 4 had lower thresholds; there is no evidence from this analysis that the nonparalel items should be dropped from the PCL:SV. Table 4 Series of Analyses to Reach Constrained Model COOKE, MICHIE, HART, AND HARE Comparison of Part I and Part 2 Items Items from Part | of the PCL:SV had, on average, higher band +b parameters than items from Part 2, U (6, 6) = 0.0, p < 001; U (6,6) = 20,p = 01}: there was no difference between items from, Parts | and 2 with respect to the a parameter, U (6, 6) = 12.5, ns. These findings are consistent with earlier analyses of the PCL-R, (Cooke & Michie, 1997, in press), which found that items loading con Factor 1 had'a higher threshold than did items loading on Factor 2. Taken together, the findings suggest that interpersonal and affective features of psychopathy have a higher threshold than. do impulsive and antisocial behavioral features: however, the interpersonal and affective features are no more discriminating, Comparison of PCL:SV and PCL-R Latent Traits For 244 participants, there were four available estimates of the latent trait underlying the scales: @acy.sy. the estimate from the item response theory analysis of PCL:SV items; Tp. ys the total, score (sum of item scores) from the PCL:SV; Ope, the estimate from the item response theory analysis of PCL-R items; and Tce» the total score (sum of item scores) from the PCL-R. The correlations among these four estimates were very high (all r = 89), and the associations were highly linea. Ifthe PCL:SV and PCL-R both measure the same latent tri then the values of @rcy.sy Nd Qrci-p for @ given participant should be the same within sampling eror. A1 value was calculated for each individual. The mean of these values was t= 4.93 with 243 degrees of freedom p <.001, which is significant suggesting there is some difference in the traits measured by the PCL:SV and, PCL-R items. To investigate this difference further, we drew a regression line and 95% confidence limits on a scaterplot of the Analysis ) + constants imposed quate equivalent PCL:SV and PCL-R items Superficial Amalysis 1+ Grandiose ‘Analysis 2 + Poor behavior controls Analysis 3+ imesponsible ‘Analysis & + Doesn't sept responsibility Atalysis $+ Lacks remorse quate PCL-R items ‘Analysis 5+ Juvenile delinquency & early behavior problems | [Analysis 7 + Nood for stimulation & impulsivity ‘Analysis 8 + Conning & psbologiea! iin Analysis 8 + Revocation & criminal vesaity -Equate PCL:SV item and? PCL-R items “Analysis 8 + Adolescent antisocial behavior Analysis 8 + Impasiity Equate PCL:SV iter and Ist PCL-R item “Ansiysis 12 + Manipulative Analysis 13+ Lacks goals Nowe af ‘ah analysis: is G? for the base (ie. unconstaines) @ a Probability 38070 sgor0 0. 3 10 Sars os 6 968 BBs 65 5 659 BRITO wo 861266 8808 RIS RS x 1077 eR? 42 RAS 10? sss sk Bass 2652 as Riso 40 28g x10 sas 41828136 10 8.98.1 te ga x 0 8860 02k 0 ssr2 ons SMe 98307 numberof item parameters constrained tobe equal: Gi G? (i.e, ~2 lg likelinoed) fr the feasts. PCLSV lre Psychopathy Checklist: Screening Version: PCL-R = Hare Psychopathy Checklist Revise, IRT ANALYSIS OF THE PCL:SV Table 5 Comparison of Hare Psychopathy Checklist: Screening Version (PCL:SV) and Revised (PCL-R) Items: Constrained Model PeLsv PCL-R em a by bs em e by ®, Superficial 18 -02 16 Superficial 1s ~02 16 Grandiose 16 =02 13 Grandiose ee 3 Deceit 7 04 12 Conning/maniputative nn re) Pathological lying a 00, i Lacks remorse a4 -05 06 Lack of remorse 25 -09 03 Tacks empathy 28 05 Mt Callouslack of empathy is 05 Ow ‘Shallow affect a4 Poneto Doesn't accept responsibilty 22 07 0.6 Faire to accept responsiblity 220-07 Impulsive 20 ou 03 Impulsiviyy 20 =u 03 [Need for stimulation 20 =u 03 Poor behavior controls 16 ~08 —0S——_—Poor behavior controls 16 “0s 0s Lacks goals 1s 1204 Lack of goals 1s “12 Parasitic este 21 0s 08 Irresponsible 2 -10 04 rresponsibility 24 “10 ad Adolescent antisocial behavior ta 08 08 Juvenile dlingueney 14 00 1a Early behavior problems 4 00 18 Adalt antisocial behavior 2s “08 03——_—_Revocation of conditional release 1 -04 os Criminal versatility 20 o1 09 ot. by are the thresholds for Py and P, respectively ‘estimates of trait from the PCL:SV items against the PCL-R items as shown in Figure 2 The plot demonstrates that the relationship between the (wo estimates of trait is well described by the linear relationship. This means thatthe estimate of tat from the PCL:SV items is @ good predictor ofthe estimate of tat from the PCL-R items and thatthe prediction is equally good across the entire range of the trait Theta From PCL:SV Items —— Fitted regression tine Items that have been consrtned to ave identical parameters are in boldface. ais he slope ofthe logistic curve atthe pont of infection 5 and According to regression analysis, the relationship between Orce.sv and Byer can be expressed as follows: Greusy = 0.130 + 0.9636 pc. a The regression analysis indicated that the association between the two latent traits was highly significant, F(1, 242) = 1,100.0, = 91, p < 001. Although the association was linea in nature, 95% confidence limits for predicted values ° Theta From PCL-R items Figwe 2 Seater plot of theta values from Screening Version of the Psychopathy Checklst—Revised {PCL'SV) items versus theta values from PCL-R items, The fied regression line wth 95% confidence limits for prediction i pote, 10 (COOKE, MICHIE, HART, AND HARE Table 6 Information Functions at Various Levels of the Trait Information at varios levels of @ 29-13-10 30 «5290 4808328 16 6k -05 00 0501S 20 Bo BS 11 278658 4 58 166 SRS 79 212 3s Note. PCL:SV = Hare Psychopathy Checklist: Screening Version; PCL-R = Hare Psychopathy Checklist Revised note that the intercept is significantly different from 0, 1(242) = 5.0, p <.001. The slope does not differ significantly from 1, (242) = 1.28, ns. These findings suggest that the PCL:SV mea- sures a wait identical to that measured by the PCL-R with the addition of a small constant, However, the small magnitude of the difference and the high proportion of shared variance suggest the difference between the scales is not psychologically meaningful, Thus, it appears that the PCL:SV and PCL-R are metrically equivalent, but that they have different origins (by approxi- rately 0.1 standardized units, ic., equivalent to 1 ina PCL-R total score), Put another way, the two tests have the same units of ‘measurement but slightly different zero points? In fact, as the estimates of trait are discrete, the difference may be due to the fact that the possible estimates of trait from the PCL:SV items are different from those from the PCL-R items. We conclude thatthe PCL:SV can be considered # good measure ofthe intensity of the trait that previously has been estimated using the PCL-R. Comparison of PCL:SV and PCL-R Information Functions ‘The information provided by a test or item is the inverse of the square of its standard error of estimate (Nunnally & Bernstein, 1994), Test or item information can be considered the item re- sponse theory equivalent of reliability in classical test theory However, unlike CTT indexes of reliability, information functions provide estimates concerning precision of measurement at various Tevels of the latent trait. For diagnostic tests such as the PCL:SV and PCL-R, information should he a maximum around the diag- nostic cutoff, Test information at various levels of # is presented in Table 6. [As would be expected from the difference in the length of the tests, the PCL-R provides more information than docs the PCL:SV at all levels of the latent trait. Around the diagnostic cutoff (ap- proximately @ = 1.0 on both tests; see below), the ratio of infor- mation is about 1.2:1, Thus, an increase in test length of 66% results in only a 20% increase in precision of measurement, For low levels of the tat, the ratio is much higher. Ths later finding suggests that the PCL-R yields more precise measurements than does the PCL:SV at low levels of 0, a finding that is somewhat surprising given thatthe PCL:SV was intended for use in nonfo- rensic populations (where, presumably, levels of @ are quite low), ‘whereas the PCL-R was not? Item information functions are presented in Table 7. The PCL'SV and PCL-R items that had identical parameters also had identical information functions; thus, information functions for PCL-R items are presented only for the four nonparallel PCL:SV Inspection of Table 7 reveals that PCL:SV Part! items provided the most information at the highest levels of the latent trait, ‘whereas Part 2 items provided the most information at low levels of the trait. Among Part 1 items, Items 4 (lacks remorse) and 5 (lacks empathy) provided the greatest information. Among Part 2 items, Items 12 (adult antisocial behavior) and 10 (responsible) provided the most information, ‘There are some noteworthy differences between the information functions of corresponding PCL:SV and PCL-R items; some of these differences are illustrated in Figure 3. For example, PCL:SV Tem 5 (lacks empathy) generally carried more information than ‘one ofthe PCL-R items from which it was derived, thats, shallow affect, but less information than the other item from which it was derived, that is, callous/lack of empathy. However, around the diagnostic cutoff (ie, around 1.0 the information provided by items lacks empathy and callous/lack of empathy were essentially identical). The PCL:SV item, adolescent antisocial behavior, per- formed no better oF no worse than the equivalent PCL-R items (ie, juvenile delinquency and early behavior problems). By way of contrast, the PCL:SV item, adult antisocial behavior, provided substantially more information than ether ofthe two PCL-R items. from which it was derived (ie, criminal versatility and revocation ‘of conditional release) Interesting to note, these information analyses provided strong. ‘support for retaining the nonparalle items inthe PCL:SV. Three of the four PCL:SV items that carried the greatest information were ‘nonparallel items; and all four nonparallel items carted as much information as did their PCL-R counterparts (albeit at different trait levels) Diagnostic Efficiency of the PCL:SV On the PCL-R, a cutoff score of =30 is used to classify individuals as psychopaths (Hare, 1991). Many researchers also consider those seoring =21 (but below 30) to constitute @ group The relation between the PCLSV and PCL-R may be considered analogous to that between the Celsius and Kelvin temperstute sles, *Icshould be kepcin mind, however, that some PCL-R items ae scored ‘on the bass of formal criminal records. Ths, even though the PCL-R appears to have good precision a low levels of the latent tt in these analyses, fr practical reasons it may be dificult or impossible to admin- ister ouside of forensic senings. IRT ANALYSIS OF THE PCL:SV u Table 7 stem Information Functions Information > 100 a various levels of @ -o5 00 0s S20 em -20 15-10 Supertic Boom @ 3 2 6 of 8 3 Grandiose ery eer Devetfl D3 © 7% 7 7% MBO PCL-R pathological lying Okt Lacks remorse 4724 00a wk PCL-R lack of remorse or guilt 39 TS} SS Lecks empathy Pata PCL-R callous/ack of empathy 6 36 37 tw PCL-R shallow afect 3 68 te 10S 705 Doesnt accept responsibility PC ee i ee ye a Impulsive 32 8 Shwe w Poor behavior controls 7 4% & 7B mM mM 3 OD Lacks goals a % 8 8 6 6 9 2 PCL-R parasitic lifestyle 2 323100085 Irresponsible “ow? SB Adolescent antisocial behavior % 38 SSS SSH PCL-R juvenile delinquency and caly behavior problems 10 Tr oe rs Adult antisocial behavior 9 tee? 502 PCL-R revocation of conditional release 65 OTS PCL-R criminal versatility 4 1 1m om oS Note: PCL-R = Hare Psychopathy Checklist—Revised with mixed or moderate psychopathic Features (ie., possible psy- chopaths). The PCL:SV manual recommends a cutoff of =18 to classify individuals as psychopaths and =13 10 classify them as possibly psychopathic and in need of further evaluation with the 20-item PCL-R (Hart et a, 1995). Given the strong linear rela- tionships between latent trait and test score on the PCL:SV and PCL-R, its possible to check the appropriateness of the recom- mended PCL:SV cutoff using Equation 1 and the following re- ‘gression equations: Preva = ~1.T1 + 0.092 here @ Trcusy = 12.30 + 6.100pe1:sv e According to these equations, a cutoff of =30 on the PCL-R corresponds with =20 on the PCL:SV, and a cutoff of =21 on the PCL-R corresponds with =15 on the PCL:SV. Thus, according to these analyses, the recommended PCL:SV cutoffs both are about 2 points too low. However, this is not panicularly problematic if the PCL:SV is used as a tue screening test, as the probability of false-negative errors is very low and as false-positive errors would bbe detected in follow-up evaluation with the PCL-R.* Discussion With respect to the two general questions posed in the Into:

You might also like