Printer Friendly

Psychometric prediction of penitentiary recidivism/Prediccion psicometrica de la reincidencia penitenciaria.

Rehabilitative interventions performed in prison settings are designed to reduce criminal incidence rates, but results so far have been disappointing (Medina, 2013). Criminal behavior is complex, has multiple causes, and is subject to legislative changes. With the exception of some general intelligence and personality characteristics (e.g., Chico, 1997; Pelechano, 2008; Sanchez-Teruel & Robles-Bello, 2013), it has not been possible to identify the profile of the delinquent. Personality scales have failed to make highly accurate predictions about recidivism. Hence, for the past few decades, actuarial scales have been used, based on a mechanical combination of empirically validated risk factors (Andres & Echeburua, 2010; Brouillette-Alarie, Proulx, & Benbouriche, 2013). Thus, Luque, Ferrer, and Capdevila (2005), after a 4.5 year follow-up of a sample of former prisoners in 1997, elaborated a logistic regression equation with criminal and sociodemographic variables. However, the predictive power of the true positives (the people who actually relapsed), was 23.4%, a percentage below the level of accuracy by chance (37.4%).

Currently, there are more than 100 actuarial prediction scales, most of them referring to violent contexts and specific behaviors. They assign a risk factor weight, both "static" (e.g., unmodifiable) and "dynamic" (e.g., changeable), that correlates with repeated entry in prison (e.g., age, impulsivity, hostility, etc.). As limitations, it should be pointed out that these scales require specific training, they take a long time to be completed, and they have a predominance of items related to non-modifiable variables (e.g., gender). Several scales can be highlighted, among them: the Sexual Violence Risk-20 (Boer, Hart, Kropp, & Webster, 1997) scale for sex offenders; The Spousal Assault Risk Assessment (Kropp & Hart, 2000); the Escala de Prediccion del Riesgo de Violencia Grave (Scale for Predicting the Risk of Serious Violence against the Partner Reviewed; Echeburua, Amor, Loinaz, & De Corral, 2010), which tries to predict recidivism against the partner; and the Assessing Risk for Violence (Webster, Douglas, Eaves, & Hart, 1997), used to predict generalized violence. The meta-analysis by Singh, Grann, and Fazel (2011) showed that the greatest predictive power reached by one of these scales, and measured by the Area Under the Curve (AUC), was 0.78 for the median and 0.71-0.83 for the interquartile range. A more recent meta-analysis by these authors (Fazel, Singh, Dol, & Grann, 2012) found a median of AUC = 0.72 for violent crimes and an interquartile range of 0.68-0.78. The values for "general recidivism" were lower, 0.66 and 0.58-0.67, respectively.

In Spain, Grana et al. (2012) used the Inventario de Factores de Riesgo e Intervencion en Prisiones (Inventory of Risk Factors and Intervention in Prisons; an adaption of Level of Service Inventory Revised, Andrews & Bonta, 1995) to predict recidivism in retrospect (N = 811). The AUC values for violent and general recidivism were 0.81 and 0.77, respectively.

The most recent attempt is the RISCANVI scale, implemented in Catalonia (Spain) in 2009. A retrospective study by Nguyen, Arbach, and Andres-Pueyo (2011) achieved an AUC value of 0.64 for violent penitentiary recidivism, commonly found in other studies (Fazel et al., 2012; Singh et al., 2011), but too low to make individual prognoses (Martinez, 2014). In another prospective study by Capdevila et al. (2015), the RISCANVI was applied to predict violent recidivism in 684 former prisoners in 2010. Although reaching a sensitivity of 77% and a specificity of 57.26%, the authors did not offer the AUC or the positive predictive value (percentage of real violent recidivists predicted compared to the total number of violent recidivists predicted). However, this percentage was obtained from table 38 (Capdevila et al., 2015; p.151): 17.94%, compared to a base violent crime rate of 29.4%, that is, 11.46% below chance. In addition, the data point out that for every 2 future violent recidivists correctly identified, another 9 will be erroneously classified who would not be. Despite these data, the authors concluded that "the tool performed quite well in predicting risk in individuals who would actually relapse (77.15%), and it was acceptable in classifying as low risk the subjects who actually would not relapse (57.26%)." (Capdevila et al., 2015, p. 237).

The violent criminal typology shows a low frequency with regard to general recidivism. For example, in Catalonia (Spain), this frequency ranges between 16.5% and 29.4% (Capdevila et al., 2015; Luque et al., 2005). For this reason, sensitivity is usually good, but not specificity, and the probability of including fake positives is a serious and unresolved problem. Indeed, authors such as Martinez (2014), after carrying out an excellent review, indicated the dangers of trusting actuarial scales too much.

Considering these criticisms and limitations of actuarial scales, the objective of this study is to prove that the evaluation of the personality traits can be useful to predict delinquent recidivism, without the need for other risk factors, "ad hoc" constructs (e.g., "criminogenic needs" of Andrews & Bonta, 1995), or other specific scales, such as actuarial scales. To achieve this, instead of using the usual personality trait subscales, elaborated on the basis of general or clinical populations, this study starts from the basic information provided by the items that make up the psychometric instrument. Specifically, it is hypothesized that the items that make up the Cuestionario de Personalidad Situacional (CPS; Inventory of Risk Factors and Intervention in Prisons; Fernandez, Seisdedos, & Mielgo, 1998) will make it possible to predict penitentiary recidivism in a broad sample of male inmates.

Method

Participants

The sample was composed of two groups of males: inmates (n = 1116) and general population (n = 1700). The inmates (M = 36.27 years old; SD = 9.64) were from 4 prisons in Catalonia (Spain) (97% from the Tarragona Prison). In the total sample, 138 participants had no prison records and were serving their sentences at the time of the study. The rest of the sample (n = 978) was divided into two groups according to the following definitions:

a) Recidivist (n = 568) (M = 35.96 years old; SD = 9.24). People who have served a sentence because of committing a crime and re-enter prison again for some other reason. This includes committing a new crime while the inmate is serving the sentence, escapes, and remand prisoners, who are released and re-enter again for another criminal case subsequent to the initial entry in prison.

b) Non-recidivist (n = 410) (M = 36.72 years old; SD = 10.16). People who enter prison because of some crime, are released, and do not re-enter prison due to a criminal case different from the initial one.

The control group was composed of a sample of male participants from different selection processes between 2007 and 2014, provided by TEA editions. They were randomly chosen and stratified by age (M = 29.9 years old; SD = 9.35) and nationality, Spain (n = 1200) and 14 Latin American countries (n = 500). There was no information about their educational level, but the different selection processes ensured the inclusion of different educational levels, from primary to university levels.

Instruments

Cuestionario de Personalidad Situacional (CPS; Fernandez et al., 1998). It is composed of 223 dichotomous items typified in a sample of 39,641 Spanish people and grouped in 15 personality variables (emotional stability, anxiety, self-concept, effectiveness, confidence, independence, dominance, cognitive control, sociability, social adjustment, aggressiveness, tolerance, social intelligence, integrity and leadership), 3 measures of validity (sincerity, social desirability, and control of answers), and 5 second-order factors (adjustment, leadership, independence, consensus and extraversion). The interquartile range and the median of the reliability of the scales were: coefficient alpha (0.587; 0.83) and 0.725, respectively; test-retest (0.89; 0.91) median: 0.89. Regarding validity, there is a wide range of correlations, according to the contrasted scales of two psychometric instruments: Questionnaire of Personality (Eysenck & Eysenck, 1994) with a range: 0.77; -0.76, and the Clinical Analysis Questionnaire (Krug, 1994), with a range: 0.64; -0.58. The CPS questionnaire was chosen for its ability to identify conflictive inmates and predict regression in the penitentiary treatment (Raya, Villacorta, & Medina, 2008), and its double typification, in a Spanish general population and a penitentiary population (Medina, 2013).

In addition, measurements were obtained for the following variables (see Table 1): age (in years), educational level (elementary, secondary, mid-level, and higher), 8 criminal categories, and nationality (Spanish and non-Spanish). Penitentiary trajectory was added because it is strongly associated with recidivism, and it is the variation of the penitentiary treatment degree in the initial classification. It was dichotomous, operationalized in two categories, no antecedents of degree regression or antecedents of degree regression.

Procedure

The inmates filled out the CPS in the real context of the prison over a period of 11 years (from 30/04/2004 to 31/06/2015). Regarding the recidivist inmates, the assessment was carried out before and after the crime was committed. Regarding the non-recidivist inmates, after checking the absence of a criminal record, the follow-up was carried out at least one year after the final release (M = 1529 days; SD = 695; range 368-3882). In Catalonia, the majority of prisoners who served a first conviction did not relapse, but it was noted that the ones who did so relapsed in an average of 359.25 and 6378 days (Capdevila et al., 2015; Luque et al., 2005).

Data analysis

The statistical analyses were performed using SPSS 15.0 in the following sequence:

1) Selection of the CPS items with a discriminative index > 0.1 in the penitentiary population (N = 1116). 209 items were obtained.

2) These items were introduced together as predictor variables in a logistic regression equation.

3) A score for each participant was calculated using the weighting coefficients of the logistic regression equation.

This new synthetic variable, called "recidivism209", was used to predict recidivism in an actuarial way. It means that an a posteriori probability calculation was made, once it had been verified which participants were recidivist and nonrecidivist.

4) An ANOVA was performed to validate the hypothesis of the predicted model: the predictor variable ("recidivism209") should discriminate between the recidivist and non-recidivist inmates. Simultaneously, there should be no differences between the non-recidivists (rehabilitated) and the control group (general population), or within the control group based on relevant characteristics, such as being Spanish or not.

5) The score on the new predictor variable, using the coordinates of the ROC curve, makes it possible to individually calculate the sensitivity and specificity. These were summarized using percentiles.

Results

The indicators of the logistic regression equation were: [chi square](209, 978) = 409.51, p<.001; test of Hosmer and Lemeshow: [[chi square].sub.(8)] = 2.907, p = .93; Nagelkerke's [R.sup.2] = 0.46. The match-accuracy is shown in Table 2, and the B coefficients in Table 3.

The discriminative capacity of the ROC curve of the predictor variable is shown in Figure 1. The corresponding statistics are: AUC = 0.85, p<.001, Se = 0.012, 95% CI [0.826, 0.873].

An example of the process for calculating the score for "recivism209" using a participant's CPS answers (e.g., item1: true; item2: false, (...), item233: false) is shown: 1) Score of the new variable using the coefficients from Table 3 is calculated. The "true" answer weight 1 and the "false" answer weight 2: -0.478.1+ 0.313.2 + (...) + 0.258.2 = 2.116). The values of the ROC curve coordinates. The value 2.116 is close to the percentile distribution (see Table 4). The closest score is Percentile = 77, which means that 77% of the participants have a score equal to or below 2.116 on "recidivism209". The sensitivity is 0.359, which means that 35.9% of the recidivist participants have a value above 2.116. The probability of obtaining a false positive is shown in column "1-specificity" of Table 4. The value is 0.034, and it is the probability of being recidivist when this is not true. In this example, this participant is probably going to be a recidivist inmate.

Validation of the predictive model. All the analyses supported the expected predictions: recidivist vs. non-recidivist, F(1, 976) = 496.14, p < .001; [[eta].sup.2] = 0.34; non-recidivist vs. control group, F(1, 2108) = 1.22, p = .269; [[eta].sup.2] = 0.001. The non-significant difference between Spanish and non-Spanish participants within the control group was also supported, F(1, 1698) = 0.52; p = .47; [[eta].sup.2] = 0.001. The descriptive statistics for the variable "recidivism209" are shown in Table 5.

Discussion

This study demonstrates that a "general" personality questionnaire may help to predict something as complex as penitentiary recidivism. The selected items only meet a minimum requirement: having a discrimination index > 0.1. More than 99% of them are dynamic (e.g., they refer to editable variables), and they do not make a direct reference to any risk behavior (e.g., drug addiction) or criminal behavior. Furthermore, a broad, heterogeneous sample was used, higher than the scale average of any study included in the cited meta-analysis. The predictive precision level found has not often been surpassed by other scales in their respective specific areas. Some items have a weight that is as much as 1100 times greater than others, which would help to group them based on factors or traits, in order to achieve a more efficient predictive scale.

The indicators of static or unmodifiable risks of the current actuarial scales reduce the scope of action of prison rehabilitation. Thus, criminal records and age have great predictive power (Molleda, Rodriguez, Perez, Sanchez, & Ovejero, 2013), especially in violent recidivism (Grana et al., 2012; Nguyen et al., 2011), but not many possible actions exist to change them. This study points out the relevance of variables that are susceptible to change.

[FIGURE 1 OMITTED]

Another important problem is that the predictive studies of recidivism often confuse penitentiary recidivism (return to prison for committing a new crime) with criminal withdrawal (abandonment of the criminal career throughout the life cycle). Therefore, some inmates are considered "non-recidivist" when they are not, because although they have not entered prison again during the follow-up period established, they have a criminal record. Although recidivism and withdrawal are related, they are not equal constructs: a person does not cease to be a recidivist because he/she spends a period of time without returning to crime. This confusion interferes in the study of psychological dimensions that explain criminal behavior. Undoubtedly, the intimidating value of sending recidivists to prison again is lower than in the case of non-recidivists. What this study shows is that their personality profiles are probably different, providing greater predictive and rehabilitative ability.

Personality scales are constructed by grouping characteristics of items that are applied to standard or clinical samples. Thus, the alleged patterns of a possible "criminal personality" are hard to detect under the specific weight of the prison population compared to the general population. These factors would reflect personal and social functioning, as a midpoint between the static and dynamic factors (Grana et al., 2013), to explain and predict recidivism and criminal withdrawal. These variables would indicate a balance between the intimidating values of imprisonment (sensitivity to punishment) and the incentive to achieve objectives by breaching the penal code (sensitivity to reward).

Studies that attempt to analyze and predict criminal recidivism have to deal with multiple biases (e.g., offenses and imprisonment backgrounds not registered in the database used as a source) because detecting all the crimes is impossible. Penitentiary recidivism as an indicator of delinquency is quite restrictive, in contrast to other possibilities (e.g., police and judicial recidivism), but it is also the most visible, easy to measure, and modifiable.

Moreover, another limitation of this study is that the minimum follow-up period was limited to one year, in order to not excessively reduce the sample. Future studies could use minimum follow-up periods of two years or more. In addition, this study has been limited to the territorial scope of Catalonia (Spain). It would be necessary to perform replications in other geographical areas, which would reduce many sources of error in the detection of itinerant criminal recidivism throughout the national territory.

Another weakness is that the prison sample used was not selected randomly, but rather based on its accessibility. Nonetheless, the reader can evaluate its representativeness by contrasting the relations between recidivism and the descriptive variables (penitentiary trajectory, education level, and nationality), and comparing the estimations obtained in other studies.

In conclusion, this study indicates that the CPS is a useful instrument for making predictions about penitentiary recidivism. The broad scope of implementation achieved in the male population (any criminal category, age, education level and nationality) facilitates its replicability in future studies and gives an important role to the personality construct in explaining criminal behavior, beyond other factors, both static and dynamic, contemplated in different criminological theories (e.g., Andrews & Bonta, 1995) and rehabilitation programs.

In the future, in addition to replicating the predictive capacity of the proposed items, it would be interesting to define their content, which traits are grouped together and which ones are the best predictors, and their psychometric characteristics. All of this information is important in designing more effective prison rehabilitation strategies.

doi: 10.7334/psicothema2015.269

References

Andres-Pueyo, A., & Echeburua, E. (2010). Assessment of the risk of violence: Available instruments and indications for application. Psicothema, 22(3), 403-409.

Andres-Pueyo, A., Lopez, S., & Alvarez, E. (2008). Valoracion del riesgo de violencia contra la pareja por medio de la SARA [Assessment of the risk of violence against the partner through the SARA]. Papeles del Psicologo, 29(1), 107-122.

Andrews, D. A., & Bonta, J. (1995). The level of service inventory-revised. Toronto, Canada: Multi-Health Systems.

Boer, D. P, Hart, S., Kropp, P. R., & Webster, Ch. D. (1997). Sexual Risk Violence-20. Lutz, Florida: Psychological Assessment Resources.

Brouillette-Alarie, S., Proulx, J., & Benbouriche, M. (2013). Mieux saisir les outils actuariels: Une analyse factorielle de la Statique-99R [A better understanding of the actuarial tools: A factor analysis of Static99R]. Criminologie, 46(1), 199-219.

Capdevila, M., Blanch, M., Ferrer, M., Andres, A., Framis, B., Comas, B., ..., Mora, J. (2015). Tasa de reincidencia penitenciaria 2014 [Prison recidivism rate 2014]. Barcelona, Spain: Centre de Estudis Juridics i Formacio Especialitzada.

Chico, E. (1997). The invariance in the factorial structure of the Raven in groups of criminals and non-criminals. Psicothema, 9(1), 47-55.

Echeburua, E., Amor. P. J., Loinaz, I., & De Corral, P. (2010). Scale for the prediction of the risk of serious violence against the partner, Reviewed EPV-R]. Psicothema, 22(4), 1054-1060.

Eysenck, H. J., & Eysenck, S. B. G. (1994). Manual of the Eysenck Personality Questionnaire. California: EdITS/Educational and Industrial Testing Service.

Fazel, S., Singh, J. P., Doll, H., & Grann, M. (2012). Use of risk assessment instruments to predict violence and antisocial behaviour in 73 samples involving 24827 people: Systematic review and meta-analysis. British Medical Journal, 345(e4692), 1-12.

Fernandez-Seara, J. L., Seisdedos, N., & Mielgo, M. (1998). CPS, Cuestionario de Personalidad Situacional [CPS, Situational Personality Questionnaire]. Madrid, Spain: TEA Ediciones.

Grana, J., Andreu, J., Silva, T., Pozuelo, F., Ruiz, A., Almeida, M., ..., Visdomine, J. (2012). Evaluacion del riesgo delictivo en Espana [Evaluation of the crime risk in Spain]. Madrid, Spain: Ministerio del Interior, Secretaria General de Prisiones.

Kropp, P. R., & Hart, S. D. (2000). The Spousal Assault Risk Assessment (S.A.R.A.) Guide: Reliability and validity in adult male offenders. Law and Human Behavior, 24(1), 101-118.

Krug, S. E. (1994). CAQ. Cuestionario de Analisis Clinico (3a edicion revisada) [Clinical analysis questionnaire, 3rd reviewed edition]. Madrid, Spain: TEA Ediciones.

Luque, E., Ferrer, M., & Capdevila, M. (2005). La reincidencia penitenciaria a Catalunya (1997-2002) [Penitentiary recividism in Catalunya 1997-2002]. Barcelona, Spain: Centre d'Estudis Juridics i Formacio Especialitzada.

Martinez, M. L. (2014). La incertidumbre de los pronosticos de peligrosidad: consecuencias para la dogmatica de las medidas de seguridad [The uncertainty of forecasts of dangerousness: consequences for dogmatic security measures]. InDret, 2, 1-77.

Medina, P. M. (2013). Evaluacion experimental de la eficacia de los programas psicologicos de tratamiento penitenciario [Experimental evaluation of the effectiveness of psychological programs for prison treatment]. Madrid, Spain: Ministerio del Interior, Secretaria General Tecnica.

Molleda, C. B., Rodriguez, F. J., Moral, M. D. L. V., Perez, B., & Ovejero, A. (2013). Comportamiento delictivo reincidente. Analisis diferencial de la variable edad [Recidivist criminal behavior. Differential Analysis of the age variable]. Interamerican Journal of Psychology, 46(3), 365-374.

Nguyen, T., Arbach, K., & Andres. A. (2011). Factores de riesgo de la reincidencia violenta en poblacion penitenciaria [Risk factors of violent recidivism in a prison population]. Revista de Derecho Penal y Criminologia, 3(6), 273-293.

Pelechano, V. (2008). Delincuencia, personalidad y psicopatologia [Delinquency, personality and psychopathology]. Analisis y Modificacion de Vonducta, 34(150-151), 13-66.

Raya, D., Villacorta, E., & Medina, P. (2008). Validacio creuada en poblacio penitenciaria de criteris psicometrics i tecnics per la prediccio de conductes adaptatives i factors de risc [Cross-validation in a prison population of psychometric and technical criteria for predicting adaptive behaviors and risk factors]. Barcelona, Espana: Centre d'Estudis Juridics i Formacio Especialitzada.

Sanchez-Teruel, D., & Robles-Bello, M. (2013). Model "Big Five" personality and criminal behavior. International Journal of Psychological Research, 6(1), 102-109.

Singh, J. P, Grann, M., & Fazel, S. (2011). A comparative study of risk assessment tools: A systematic review and meta-regression analysis of 68 studies involving 25,980 participants. Clinical Psychology Review, 31(3), 499-513.

Webster, C. D., Douglas, K. S., Eaves, D., & Hart, S. (1997). HCR-20: Assessing risk for violence (-version 2). Burnaby, British Columbia: Simon Fraser University.

Pedro Manuel Medina (1) Garcia and Rosa Maria Banos Rivera (2)

(1) Centro Penitenciario de Tarragona and (2) Universidad de Valencia

Received: October 14, 2015 * Accepted: February 5, 2016

Corresponding author: Pedro Manuel Medina Garcia

Centro Penitenciario de Tarragona

C/Violant d'Hongria, 39 esc. 2 piso 2 puerta 2

43007 Tarragona (Spain)

e-mail: pmmg8@yahoo.es
Table 1

Frequencies distribution by educational level, criminal
category, nationality and variation in the penitentiary treatment
degree (regression of degree)

                           Recidivists     Non-recidivists   Total

                            N      %       N        %        N      %

Educational level
Elementary                 227   40.0     137      33.4     364   37.2
Secondary                  304   53.5     197      48.0     501   51.2
Mid-level                  13     2.3      32      7.8      45     4.6
Higher                     18     3.2      21      5.1      39     4.0
No data                     6     1.1      23      5.6      29     3.0
Criminal Category
Against people             106   18.7      68      16.6     174   17.8
Sexual assault             27     4.8      33      8.0      60     6.1
Domestic violence          62    10.9      63      15.4     125   12,8
Theft                      282   49.6      81      19.8     363   37.1
Economic crimes            22     3.9      24      5.9      46     4.7
Against public health      45     7.9      99      24.1     144   14.7
Against traffic safety     16     2.8      31      7.6      47     4.8
Others                      8     1.4      11      2.7      19     1.9
Nationality
Spanish                    496   87.3     324      79.0     820   83.8
Non-Spanish                72    12.7      86      21.0     158   16.2
Regression of degree
Regression of degree       248   43.7      56      13.7     304   31.1
Non-regression of degree   320   56.3     354      86.3     674   68.9
Total                      568   100.0    410     100.0     978   100.0

Note: The total sum of rates varies one decimal point due to the
approximation error

Table 2

Classification of the logistic regression equation

                              Predicted     Match-
                              recidivism    accuracy

                              No      Yes

Observed tecidivism     No    325     85    79.3%
                        Yes   137     431   75.9%
Global match-accuracy             77.3%

Table 3

B Weighting coefficients of the scores on the CPS

Variable     B

Item1      -0.478
Item2       0.313
Item3      -0.537
Item4       0.083
Item5      -0.304
Item6      -0.321
Item7      -0.213
Item8      -0.084
Item9       0.233
Item1O      0.132
Item11      0.186
Item12     -0.051
Item13      0.171
Item14      0.396
Item18      0.221
Item19     -0.04
Item21     -0.142
Item22      0.138
Item23     -0.061
Item24      0.192
Item25      0.012
Item26      0.258
Item27      0.159
Item28     -0.45
Item29     -0.194
Item30     -0.031
Item31      0.21
Item33     -0.306
Item34      0.373
Item35     -0.008
Item36      0.064
Item37     -0.135
Item38     -0.618
Item39     -0.391
Item40      0.162
Item41     -0.374
Item42      0.099
Item43      0.128
Item44     -0.216
Item46      0.008
Item47     -0.092
Item48     -0.075
Item49      0.502
Item51      0.012
Item52      0.149
Item53      0.257
Item54     -0.236
Item55     -0.072
Item56      0.122
Item57      0.412
Item58      0.404
Item59     -0.25
Item61     -0.728
Item62     -0.23
Item63     -0.016
Item64     -0.85
Item66      0.544
Item67      0.312
Item68     -0.198
Item69      0.191
Item70     -0.305
Item71     -0.449
Item72      0.804
Item74      0.2
Item77      0.491
Item79     -0.417
Item80      0.135
Item82      0.257
Item83      0.235
Item84     -0.012
Item85      0.058
Item86      0.066
Item87     -0.023
Item88     -0.25
Item89      0.137
Item90     -0.306
Item91     -0.003
Item92     -0.073
Item93     -0.332
Item94     -0.181
Item95      0.397
Item96      0.239
Item97      0.174
Item98      0.373
Item99     -0.205
Item100    -0.422
Item101    -0.428
Item102    -0.549
Item103     0.574
Item104    -0.32
Item105     0.137
Item106     0.516
Item107     0.117
Item108     0.137
Item109     0.069
Item110     0.359
Item111    -0.297
Item112     0.085
Item113     0.057
Item114     0.573
Item115     0.353
Item117    -0.158
Item118    -0.611
Item119    -0.313
Item120    -0.061
Item121    -0.439
Item122     0.341
Item123    -0.414
Item124    -0.798
Item125    -0.605
Item126     0.438
Item127    -0.095
Item128    -0.266
Item129    -0.265
Item130    -0.376
Item131    -0.361
Item132     0.282
Item133     0.011
Item134     0.173
Item135     0.098
Item136    -0.618
Item137     0.24
Item138     0.523
Item139    -0.005
Item140     0.012
Item141    -0.309
Item142     0.352
Item143     0.18
Item144     0.375
Item145     0.3
Item146     0.144
Item147     0.365
Item148     0.217
Item150     0.234
Item151     0.366
Item152    -0.337
Item153    -0.403
Item154    -0.502
Item155    -0.47
Item156     0.739
Item157     1.117
Item158    -0.061
Item159    -0.531
Item160    -0.203
Item161    -0.008
Item162     0.24
Item165    -0.383
Item166    -0.246
Item167     0.427
Item169    -0.334
Item170     0.123
Item171     0.084
Item172     0.405
Item174    -0.096
Item175     0.209
Item176     0.415
Item177    -0.575
Item178    -0.077
Item179    -0.413
Item180    -0.699
Item181    -0.269
Item182     0.427
Item183     0.77
Item186    -0.053
Item187     0.2
Item188     0.517
Item189    -0.034
Item190     0.125
Item191     0.046
Item192     0.06
Item193    -0.569
Item194    -0.29
Item195     0.284
Item196    -0.184
Item197    -0.009
Item198     0.131
Item199    -0.36
Item200     0.189
Item202    -0.492
Item203    -0.153
Item204     0.18
Item205     0.157
Item207     0.084
Item208     0.236
Item209    -0.145
Item210    -0.047
Item211     0.025
Item212    -0.261
Item213    -0.303
Item214     0.193
Item215     0.288
Item216    -0.027
Item217     0.689
Item218     0.073
Item219    -0.197
Item220     0.806
Item221    -0.068
Item222     0.299
Item223    -0.255
Item224    -0.001
Item225    -0.604
Item226    -0.291
Item227     0.486
Item228    -0.153
Item229     0.619
Item230     0.042
Item231     0.276
Item232    -0.038
Item233     0.258

Table 4

ROC curve coordinates for "recidivism209"

Score    Percentile   Sensitivity   1-Specificity

-3.600       1           1.000          0.978
-3.190       2           0.998          0.961
-2.844       3           0.998          0.939
-2.502       4           0.998          0.917
-2.379       5           0.998          0.895
-2.206       6           0.998          0.871
-2.074       7           0.995          0.854
-1.925       8           0.995          0.832
-1.801       9           0.995          0.807
-1.728       10          0.995          0.785
-1.608       11          0.989          0.768
-1.509       12          0.984          0.754
-1.445       13          0.981          0.734
-1.301       14          0.975          0.717
-1.219       15          0.972          0.695
-1.158       16          0.967          0.678
-1.075       17          0.961          0.663
-0.989       18          0.958          0.646
-0.926       19          0.956          0.622
-0.876       20          0.952          0.600
-0.834       21          0.952          0.578
-0.803       22          0.942          0.571
-0.725       23          0.935          0.556
-0.689       24          0.928          0.532
-0.631       25          0.923          0.515
-0.550       26          0.919          0.498
-0.482       27          0.910          0.488
-0.403       28          0.903          0.473
-0.339       29          0.898          0.459
-0.302       30          0.891          0.446
-0.243       31          0.887          0.424
-0.217       32          0.882          0.410
-0.178       33          0.879          0.388
-0.133       34          0.870          0.378
-0.103       35          0.864          0.359
-0.074       36          0.850          0.354
-0.008       37          0.840          0.346
0.053        38          0.833          0.334
0.114        39          0.820          0.322
0.185        40          0.820          0.298
0.221        41          0.810          0.285
0.256        42          0.801          0.276
0.302        43          0.792          0.263
0.357        44          0.783          0.251
0.392        45          0.775          0.241
0.422        46          0.768          0.227
0.473        47          0.761          0.210
0.513        48          0.750          0.200
0.596        49          0.738          0.190
0.624        50          0.725          0.183
0.687        51          0.715          0.173
0.743        52          0.704          0.166
0.773        53          0.688          0.159
0.817        54          0.676          0.154
0.896        55          0.662          0.139
0.927        56          0.653          0.129
0.971        57          0.639          0.124
1.017        58          0.627          0.112
1.113        59          0.614          0.107
1.147        60          0.599          0.107
1.209        61          0.585          0.102
1.268        62          0.572          0.098
1.339        63          0.555          0.095
1.399        64          0.540          0.088
1.453        65          0.528          0.080
1.491        66          0.518          0.073
1.520        67          0.505          0.068
1.579        68          0.493          0.063
1.633        69          0.479          0.061
1.679        70          0.463          0.056
1.753        71          0.451          0.051
1.809        72          0.437          0.044
1.877        73          0.421          0.044
1.937        74          0.405          0.044
1.975        75          0.393          0.039
2.052        76          0.375          0.037
2.104        77          0.359          0.034
2.171        78          0.345          0.032
2.261        79          0.329          0.029
2.326        80          0.315          0.024
2.369        81          0.301          0.022
2.442        82          0.287          0.017
2.562        83          0.273          0.012
2.650        84          0.254          0.010
2.785        85          0.238          0.007
2.848        86          0.222          0.007
3.013        87          0.208          0.005
3.185        88          0.192          0.005
3.282        89          0.176          0.005
3.420        90          0.157          0.005
3.539        91          0.141          0.005
3.603        92          0.129          0.000
3.721        93          0.113          0.000
3.965        94          0.093          0.000
4.114        95          0.077          0.000
4.229        96          0.062          0.000
4.456        97          0.046          0.000
4.755        98          0.030          0.000
5.159        99          0.014          0.000

Table 5

Descriptive statistics for "recidivism209"

                   n       M       SD          95% CI

Inmate group
Non-Recidivists   410    -0.576   1.434   -0.715   -.0436
Recidivists       568    1.636    1.600   1.504    1.768
Control group
Spanish           1200   -0.657   1.700
Non-Spanish       500    -0.723   1.729
Total             1700   -0.677   1.708
COPYRIGHT 2016 Colegio Oficial De Psicologos Del Principado De Asturias
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2016 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:texto en ingles
Author:Garcia, Pedro Manuel Medina; Rivera, Rosa Maria Banos
Publication:Psicothema
Date:Apr 1, 2016
Words:5280
Previous Article:The mediating role of emotional symptomatology between anticipatory fatigue and the perception of fatigue/El papel mediador de la sintomatologia...
Next Article:Spanish version of the link's perceived devaluation and discrimination scale/Version espanola de la escala perceived devaluation and discrimination...
Topics:

Terms of use | Privacy policy | Copyright © 2021 Farlex, Inc. | Feedback | For webmasters