Juveniles ROR 99 ExSum

Page 1

CJA

NEW YORK CITY CRIMINAL JUSTICE AGENCY NEW YORK CITY CRIMINAL USTICE AGENCY

Jerome E. McElroy Executive Director Mary A. Eckert, Ph.D. Associate Director for Research

RELEASE-ON-RECOGNIZANCE RECOMMENDATION SYSTEM FOR JUVENILE OFFENDERS ARRAIGNED IN NEW YORK CITY ADULT COURTS PART I. VALIDATION STUDY EXECUTIVE SUMMARY

Mary T. Phillips, Ph.D. Project Director

September 1999

52 Duane Street, New York, NY 10007

(646) 213-2500


RELEASE-ON-RECOGNIZANCE RECOMMENDATION SYSTEM FOR JUVENILE OFFENDERS ARRAIGNED IN NEW YORK CITY ADULT COURTS PART I. VALIDATION STUDY EXECUTIVE SUMMARY

Written by: Mary T. Phillips, Senior Research Analyst Marian Gewirtz, Senior Research Analyst

Systems Programmer: Dale Sealy Programmer/Analyst

Research Project Staff: Raymond Caligiure Senior Research Assistant Alisha Fenton Research Assistant Aaron Kupchick Research Assistant Vanessa Martin Research Assistant Bernice Linen-Reed Administrative Assistant

September 1999

Š 1999 NYC Criminal Justice Agency, Inc.


ACKNOWLEDGEMENTS

This work was partially funded by the Annie E. Casey Foundation as part of New York City’s participation in the Juvenile Detention Alternatives Initiative (JDAI), which had as a goal the reduction of reliance on secure detention for juveniles.

We are also grateful to the

Department of Juvenile Justice (DJJ) for supplying supplemental data used in this research and to the Office of Court Administration (OCA), whose records we used to check data in the computerized database maintained by the New York City Criminal Justice Agency (CJA). The assistance provided by the Annie E. Casey Foundation, DJJ, and OCA does not imply these organizations’ agreement with the the methods, analyses, or conclusions of this research. CJA is solely responsible for the contents of this report.


INTRODUCTION TO THE EXECUTIVE SUMMARY1 On April 28, 1996, the New York City Criminal Justice Agency (CJA) implemented a release-on recognizance (ROR) recommendation system for children under the age of 16 who are brought for prosecution in the adult courts in New York City under the State’s Juvenile Offender Law.2

The

recommendation system provides the Criminal (lower) Court3 at arraignment with a recommendation for ROR based on likelihood of failure to appear (FTA). Empirical research using a cohort of arrests in Fiscal Year (FY) 1992 (July 1, 1991, through June 30, 1992) showed that FTA for juvenile offenders (JOs) was most strongly associated with not attending school and not expecting someone at arraignment. Consequently, these two items became the basis for the JO recommendation system. Research results for this project are being issued in two separate reports. Part I, summarized here, reports the relationship between the JO release-recommendation system and release rates, and validates the system by testing the results on a cohort of 988 juveniles arrested from April 28, 1996 (the implementation date) through June 30, 1997, and subsequently arraigned in Criminal Court. In order to examine FTA, the cases were followed until completion or until June 30, 1998, whichever was earlier. The analysis was conducted on a subset of 705 JOs who were released before sentencing, thus rendering them at risk for FTA. The criterion for success was whether the recommendation system was as accurate in predicting FTA for the more recent set of arrests (after implementation) as it had been for the FY 1992 cases used in its development. Part II will examine related issues such as the timing of release and FTA; length of time it took defendants to return to court after failing to appear; and factors that may be associated with the detention of JOs who were recommended for release. In addition, the validation results will be replicated to exclude postconviction FTA data (this research included FTA through sentencing for convicted defendants) to confirm the adequacy of the model for pre-trial FTA, given the different circumstances of pre- and post-conviction releases.

1

The full report (August 1999) is available upon request from CJA. The Juvenile Offender Law, passed in 1978, provides for cases of 14- and 15-year-olds charged with certain serious felony offenses, and 13-year-olds charged with second-degree murder, to be initiated in the adult court system. In New York’s “waiver-down” system, JO cases begun in the adult court subsequently may be transferred for resolution into the Family Court based on changes in the prosecution charges or “in the interest of justice.” 3 In New York City, most criminal cases are initiated in the lower (Criminal) court beginning with the arraignment appearance. However, in New York’s two-tiered court system cases sustained for prosecution at the felony-severity level are transferred into the superior (Supreme) court for disposition. 2


ROR-Recommendation System for Juvenile Offenders, Part I: Validation Study

Executive Summary

CJA’S JUVENILE OFFENDER RECOMMENDATION SYSTEM CJA is a not-for-profit corporation that provides a variety of criminal justice services to the City of New York through a contract with the Mayor’s Office of the Criminal Justice Coordinator. One of these services consists of interviewing arrestees held for Criminal Court arraignment in order to provide prosecutors, defense counsel, and judges with a recommendation for ROR, based on background information regarding community ties. CJA staff attempt to verify some of this information by calling a third party named by the arrestee. For juveniles, the recommendation is based on school attendance, for which verification is attempted; and on whether the child expects someone (family member or friend) to come to the arraignment, for which no verification is attempted. A JO receives a positive recommendation if he or she claims to be enrolled in school either full or part time, and this response is verified. A negative recommendation is assigned if it is verified that the child is not attending school. If school status is verified, the recommendation is based on that criterion alone. If school status (attending or not attending) is not verified, the juvenile is recommended only if he or she is expecting someone at arraignment. See Figure E-1. A recommendation may be overridden by a policy consideration that excludes the arrestee from being assigned either a positive or a negative recommendation category. The four policy exclusions (which also apply to adults) are: an outstanding bench warrant; no NYSID sheet4; arrest on a murder charge; or an incomplete interview. The six JO recommendation categories are listed in the box below in the order of precedence by which they are assigned.

JUVENILE OFFENDER RECOMMENDATION CATEGORIES [J7] Juvenile Offender: No Recommendation Due To: [J7A] A. Bench Warrant Attached to NYSID [J7B] B. No NYSID Available [J7C] C. Murder Charge (125.25 or 110/125.25) [J7D] D. Interview Incomplete [J6] Juvenile Offender: Not Recommended [J5] Juvenile Offender: Recommended

4

New York State Identification (NYSID) sheets are initiated by the defendant’s fingerprints and describe his or her criminal history (also known as “RAP” sheets).

-2-


ROR-Recommendation System for Juvenile Offenders, Part I: Validation Study

Executive Summary

New York City Criminal Justice Agency Figure E-1 Juvenile Offender Recommendations: Criteria for Assigning Categories The CJA Juvenile Offender Recommendation System is based on information collected in a pre-arraignment interview. CJA staff attempt to verify some information, including school attendance, by calling a third party named by the defendant. This diagram shows how juveniles’ responses to the two criterion questions determine the recommendation category assigned.

1. Does defendant attend school part- or full-time?

Yes verified J5 Recommended

Yes or No (unverified) or Unresolved conflict [go to next question]

No verified J6 Not Recommended

2. Does defendant expect someone at arraignment?

Yes J5 Recommended

No J6 Not Recommended

A positive recommendation is assigned to juveniles who say they are in school, and this information is verified; and to juveniles whose school attendance is not verified if they are expecting someone (family or friend) at arraignment.

A negative recommendation is assigned to juveniles who say they are not in school, and this information is verified; and to juveniles whose school attendance is not verified if they are not expecting anyone (family or friend) at arraignment.

Either a positive or negative recommendation may be overriden by policy considerations. No recommendation is assigned if there is an outstanding bench warrant, there is no New York State Identification sheet containing the defendant’s criminal history, the defendant is charged with murder, or the interview was not completed.

-3-


ROR-Recommendation System for Juvenile Offenders, Part I: Validation Study

Executive Summary

RESEARCH RESULTS — HIGHLIGHTS I. Comparing the FY 1992 and FY 1997 Cohorts5 •

There were fewer JO cases in recent years. The number of JO cases arraigned in the adult Criminal Court in New York City dropped almost in half, from 1552 cases in FY 1992 to 792 in FY 1997.

Defendant and case characteristics did not change much. Although the later cohort was smaller, it resembled the earlier cohort in personal characteristics and most case characteristics and outcomes. In both groups, a JO was most likely to be 15 years old, male, black, with no prior felony convictions or open cases. Most defendants were charged with robbery, and less than half of the cases were transferred to Supreme Court. About a quarter of the cases were transferred to Family Court, with the remainder being dismissed in Criminal Court. Over 90 percent of the cases that were transferred to Supreme Court ended in a conviction, with about half of the convicted JOs receiving probation.

Release rates fell and bail amounts rose. The one striking difference between the two cohorts was in release rates: 52 percent of the FY 1992 cohort were released on recognizance (ROR’d) at Criminal Court arraignment, compared to only 38 percent of the later cohort. This is a particularly significant finding in light of the fact that the JO recommendation system had been implemented with the hope that it would result in higher release rates. Moreover, the mean bail amount rose from $6,324 to $8,598.

FTA rates6 rose slightly. The FTA rate to completion of the case for at-risk JOs was 8 percent for FY 1992 and 10 percent for FY 1997 (11% for the full 14-month 1996-1997 cohort).

5

The validation data set contained 14 months of arrests, including all of FY 1997 (July 1, 1996, through June 30, 1997). For the comparisons with the development data set of FY 1992 cases reported on this page, data for FY 1997 were extracted from the 14-month data set. No important differences were found between the 12-month FY 1997 data and the full 14 months of data. 6 This is a case-based FTA rate obtained by dividing the number of cases with a defendant who failed to appear by the total number of at-risk cases. Only one FTA per case was counted, from arraignment in Criminal Court through sentencing in Supreme Court (through final disposition, if there was no conviction), or until June 30, 1998, whichever was earlier. -4-


ROR-Recommendation System for Juvenile Offenders, Part I: Validation Study

Executive Summary

II. The Relationship between the ROR-Recommendation System for Juveniles and Release7 •

A large majority of JOs were recommended for ROR. 79 percent were recommended citywide, and 10 percent were not recommended, with wide borough differences. (The remainder were excluded from a recommendation for policy reasons.)

A JO who was recommended was more likely to be ROR’d than one who was not recommended. 42 percent of recommended JOs were ROR’d at Criminal Court arraignment, compared to 31 percent of JOs who were not recommended. This indicates that the recommendation category was related to the release decision, but not strongly.

The majority of those who were recommended were not released at arraignment. Even though being recommended did increase the likelihood of being released at arraignment in Criminal Court, it was still the case that less than half of those who were recommended were ROR’d (42%, as already noted); 4 percent made bail; 55 percent were held on bail, and less than 1 percent were remanded without bail (total exceeds 100% because of rounding). In spite of the positive relationship that was found between being recommended and being released at

arraignment in Criminal Court, we do not know if the recommendation system had much effect on the release decision. The criteria used in the recommendation system may be similar to some of the criteria traditionally used by arraignment judges in making release decisions, thus accounting for the relationship. Judges must also take into consideration a number of other factors that are not included in the CJA recommendation, such as the likelihood of conviction and the possible sentence. These data suggest that the CJA recommendation did not appear to outweigh other factors influencing the courts’ release decisions.

7

The research results reported on this and the following pages are for the full validation data set of 1996-1997 JO cases, which includes arrests from April 28, 1996, to June 30, 1997.

-5-


ROR-Recommendation System for Juvenile Offenders, Part I: Validation Study

Executive Summary

III. Evaluating How Well the ROR-Recommendation System for Juveniles Predicted FTA Evaluating the ROR-recommendation system’s ability to predict FTA was approached in a variety of ways. First, we compared FTA rates for recommended versus not-recommended JOs. Second, we examined interrelationships among recommendation, FTA, and several other factors that could affect the way recommendation and FTA were related. Third, the configuration analysis that was used to develop the recommendation system was replicated with the validation data set of 1996-1997 JO cases. Fourth, the system’s predictive accuracy was tested. Finally, a logistic regression model was estimated using the JO recommendation categories as the independent variable and FTA as the dependent variable. All of these techniques produced statistically significant results indicating that the JO recommendation system predicted FTA with a fairly high degree of accuracy. A summary of the findings from each of these approaches is provided in this section. 1. Comparing FTA Rates for Recommended and Not-Recommended Juveniles •

FTA rates were much lower for recommended JOs. A released JO who was not recommended was 3.2 times more likely to fail to appear than one who was recommended. The FTA rates for the two groups were 9 percent (for those who were recommended) and 29 percent (for those who were not recommended). This indicates that the recommendation system differentiated fairly well between juveniles with low and high risk of FTA.

2. Interrelationships Among Recommendation, FTA, and Other Factorss •

The relationship between recommendation and FTA was stronger for those released on recognizance than for those released on bail. When those who had been released on bail were excluded, a JO who was not recommended was almost four times as likely to FTA than one who had been recommended. The FTA rates for JOs who were ROR’d were 8 percent (recommended) and 31 percent (not recommended).

Release on bail did not affect the FTA rate. Release on bail had no overall effect on FTA rates, which were identical for those released on bail and for those who were ROR’d. Bail did appear to be associated with a lower FTA rate for JOs who were not recommended for release, but the number of cases was too small to draw conclusions. Bail amounts were actually higher for JOs who were recommended, so the lack of an association between release on bail and FTA for them could not be attributed to lower bail amounts.

-6-


ROR-Recommendation System for Juvenile Offenders, Part I: Validation Study

Executive Summary

High bail amounts were no more effective in discouraging FTA than low bail amounts. Among those who were released on bail, the median bail amount for cases with an FTA was $2000, compared to $1500 for those with no FTA. Furthermore, JOs who posted high bail amounts had a higher FTA rate (16%) than those who posted low bail amounts (FTA=7%).

The relationship between recommendation and FTA remained strong, controlling for charge severity. Charge severity was related to FTA: JOs charged with a Class-B felony had a higher FTA rate (13%) than those charged with a Class-C felony (FTA = 7%), which is a less severe felony class. However, the overall effect of recommendation on FTA was stronger than the effect of charge severity on FTA, and recommended JOs had a lower FTA rate than non-recommended JOs within each severity class.

3. Replication of the Configuration Analysis Configuration analysis is the technique that was used to develop the recommendation system. This analytic technique identifies a hierarchy of factors that affect the dependent variable — in this case, FTA. The procedure begins by dividing a sample into groups with different outcomes (such as FTA rates) based on one factor, then subdividing the groups again based on some other factor, and so on until no further factor can be found with an effect on the dependent variable. The factor with the strongest statistically significant effect on the outcome is used at each stage as the basis for the division. •

School attendance and expecting someone at arraignment were the strongest predictors of FTA for the FY 1992 cohort. The overall FTA rate for the at-risk JOs in the FY 1992 cohort was about 8 percent. Over 30 factors were examined to determine whether they had any effect on this rate. The factor with the strongest statistically significant relationship with FTA was school attendance. For those with verified school attendance, the FTA rate was only 6 percent, compared to 21 percent for those whose nonattendence was verified. For JOs with an unverified school status (either attending or not attending), the only additional factor that affected FTA was expecting someone at arraignment. JOs with an unverified school status who expected someone at arraignment had an FTA rate of 10 percent, compared to 16 percent for those who did not expect anyone. (Figure E-2, top) Both factors were also strong predictors of FTA for the validation (1996-97) cohort. The overall FTA rate for the at-risk JOs in the 1996-97 cohort was about 11 percent. For those with verified school attendance, the FTA rate was 9 percent, compared to 19 percent for those who were verified to be not attending school. For JOs in the 1996-97 cohort whose school attendance was not verified but who were expecting someone at arraignment, the FTA rate was 11 percent, compared to 25 percent for those who did not expect anyone. (Figure E-2, bottom)

-7-


ROR-Recommendation System for Juvenile Offenders, Part I: Validation Study

Executive Summary

New York City Criminal Justice Agency Figure E-2 Configuration Analysis: Development and Validation Results Development: FY 1992 JO Cohort FY 1992 JOs who were at risk of failure N = 1196 8.4% FTA

School Attendance No verification (yes or no), or Conflicting statements made by contact and defendant N=397

Yes Verified N=646 6% FTA (n=39)

No Verified N=39 21% FTA (n=8)

Expects Someone at Arraignment No N=93 16% FTA (n=15)

Yes N=304 10% FTA (n=30)

Total: Yes verified + Yes [combined criteria for “Recommended”] N = 950 7% FTA (n=69)

Comparing FTA rates: 17% is greater than 7% by a factor of 2.4

Total: No verified + No [combined criteria for “Not Recommended”] N = 132 17% FTA (n=23)

Validation: 1996-1997 JO Cohort 1996-1997 JOs who were at risk of failure N = 705 10.7% FTA

School Attendance Yes Verified N=382 9% FTA (n=34)

No verification (yes or no), or Conflicting statements made by contact and defendant N=266

No Verified N=21 19% FTA (n=4)

Expects Someone at Arraignment Yes N=209 11% FTA (n=22)

Total: Yes verified + Yes [combined criteria for “Recommended”] N = 591 9% FTA (n=56)

No N=57 25% FTA (n=14)

Comparing FTA rates: 23% is greater than 9% by a factor of 2.6

-8-

Total: No verified + No [combined criteria for “Not Recommended”] N = 78 23% FTA (n=18)


ROR-Recommendation System for Juvenile Offenders, Part I: Validation Study

Executive Summary

In both cohort groups, those who fulfilled the criteria for the “not recommended” category were more than twice as likely to fail to appear as those who fulfilled the criteria for the “recommended” category. Verified school attendance did not have as great an effect on FTA for 1996-1997 JOs as it did for the FY 1992 cases, and expecting someone at arraignment had a greater effect on FTA for the 1996-1997 cases, but the combined elements produced similar results for the two cohorts. The round-cornered boxes at the bottom of each part of Figure E-2 show the FTA rates for the two components combined: those who would qualify for a positive recommendation on the left, and those who would qualify for a negative recommendation on the right. (These figures include defendants who in practice would not be assigned either recommendation category because of policy exclusions.)

4. Predictive Accuracy On all measures of predictive accuracy, the validation results met or exceeded the results obtained for the development cohort.

These measures were derived from a comparison of predictions and actual

outcomes, where “recommended” was considered a prediction of no FTA, and “not recommended” was considered a prediction of FTA. When predictions were compared to outcomes, there were four possible combinations: True negatives: (TN) recommended for ROR (predicted not to fail) and did not fail; False negatives: (FN) recommended for ROR (predicted not to fail) but failed; True positives: (TP) not recommended for ROR (predicted to fail) and failed; False positives: (FP) not recommended for ROR (predicted to fail) but did not fail. •

Over 80 percent of the predictions were correct for both cohorts. If the JO recommendation system had been applied to the FY 1992 cohort, it would have resulted in a correct prediction about 84 percent of the time. The proportion of correct predictions was about 85 percent for the validation cohort.

False positives and false negatives were low for both cohorts, and more evenly distributed for the validation cohort. For the FY 1992 cohort, the errors in prediction would have included more false positives (10%) than false negatives (6%). For the validation cohort the errors in prediction were about 8 percent for each type.

For both cohorts, the system did much better in predicting no FTA than it did in predicting FTA. Because most released JOs did not fail, a prediction of non-failure was most likely to be correct even if predictions had been made randomly. Both aspects of prediction were a little more accurate for the validation cohort: 92 percent of cases with no FTA were correctly predicted, and 29 percent of cases with FTA were correctly predicted.

(Tables containing these measures are provided in Appendix A.)

-9-


ROR-Recommendation System for Juvenile Offenders, Part I: Validation Study

Executive Summary

5. Assessing the Odds of FTA The final analysis done to validate the ROR-recommendation system was estimating the odds of FTA for those who were recommended contrasted with those who were not. The statistical technique used to determine the odds was logistic regression. The model, which was estimated only for the validation data set of 1996-1997 cases, is presented in Appendix B. •

The odds of FTA more than quadrupled for juveniles who were not recommended, compared to the odds of FTA for those who were recommended. This finding was statistically significant (p=.0000). The overall model was also statistically significant at the level of p=.0001, which can be interpreted to mean that the probability is less than one in 1,000 that the relationship between recommendation and FTA could have occurred by chance.

CONCLUSIONS The JO release-recommendation system was successfully validated, as it predicted FTA for a group of defendants followed after implementation as accurately as it did for the defendants whose case histories were used to develop the system. The likelihood of FTA for most juveniles is very low, but the recommendation system identified a small group with a significantly higher likelihood of FTA. For the validation cohort of 1996-1997 arrests, errors in prediction were evenly balanced between predicting FTA when none occurred and predicting no FTA when it did occur. The overall accuracy of the system was high, with 85 percent correct predictions. However, introduction of the JO recommendation system did not halt a decline in release rates for juveniles. This is cause for concern, not only because some juveniles may be detained who could be released without much fear that they would not return, but also because detention in Criminal Court has implications for what happens later in the case. The Criminal Court release decision was strongly associated with both prosecution and release in Supreme Court. For example, a JO who was released in Criminal Court was almost certain to be released in Supreme Court, but only half of those who had been in detention throughout Criminal Court were ever released once the case went to Supreme Court. It was necessary to restrict the sample to released defendants in order to ensure that all defendants in the analysis had an opportunity to fail. The drawback is that the data can show only that released JOs who were recommended had a much lower risk of FTA than those who were not recommended. The hope that the recommendation system would encourage more use of ROR depends on the assumption that the large number of juveniles who were detained in spite of a positive recommendation were also at very low risk of

- 10 -


ROR-Recommendation System for Juvenile Offenders, Part I: Validation Study

Executive Summary

FTA. Preliminary work suggests that sample bias introduced by including only released defendants is not a problem for this research, but the issue will be addressed further in the forthcoming Part II report. Meanwhile, the current study results indicate that the ROR-recommendation system for JOs provides the courts with a valuable, objective tool — one that has now been statistically validated — for enhancing the assessment of the risk of FTA. After the completion of Part II of this study, we hope to follow up by initiating discussions with judges to improve our understanding of how they use the CJA recommendation, and what factors are contributing to the higher detention rates for juveniles that have been observed since the mid-1990s.

- 11 -


ROR-Recommendation System for Juvenile Offenders, Part I: Validation Study

Executive Summary

New York City Criminal Justice Agency Appendix A Classification Tables: Development and Validation Results Actual Outcome No FTA

FTA

Totals Total Correct predictions Total Incorrect predictions Sensitivity Specificity RIOC Base for %

Predicted Outcomes Development: FY1992 Cohort Validation: 1996-1997 Cohort J5 J6 J5 J6 Recommended Not Recommended Recommended Not Recommended Predict no FTA Predict FTA Predict no FTA Predict FTA True negatives (TN) False Positives (FP) True negatives (TN) False Positives (FP) 881 109 530 49 81.4% 10.1% 81.7% 7.6% False negatives (FN) True positives (TP) False negatives (FN) True positives (TP) 69 23 50 20 6.4% 2.1% 7.7% 3.1% 950 132 580 69 87.8% 12.2% 89.4% 10.6% TN + TP = 904 (83.5%)

TN + TP = 550 (84.7%)

FN + FP = 178 (16.5%)

FN + FP = 99 (15.3%)

25.0% 89.0% 14.1% N = 1082 at-risk cases (114 missing data on criterion variables were excluded from the at-risk sample, N=1196)

28.6% 91.5% 20.1% N = 649 at-risk cases (56 missing JO recommendation category, or with no recommendation due to exclusionary factors, were omitted from the at-risk sample, N=705)

Sensitivity: the percent of FTAs correctly predicted (TP \ FN + TP) Specificity: the percent of no FTAs correctly predicted (TN \ FP + TN) RIOC: a measure of the model’s improvement over random classification relative to the greatest improvement over random classification that would be possible given the FTA rate and the proportion who were recommended. For the calculation and interpretation of the RIOC score, see Appendix E of the full report.

- 12 -


ROR-Recommendation System for Juvenile Offenders, Part I: Validation Study

Executive Summary

New York City Criminal Justice Agency Appendix B Logistic Regression Analysis Predicting Failure to Appear1 Application of the JO Recommendation System to 1996-1997 JO Arrests AT-RISK ONLY N=6602 Independent Variable JO ROR RECOMMENDATION J5 Recommended (Reference Category) n=580 J6 Not Recommended

Logit Coefficient

Significance Level

Odds Ratio

n=69 J7A No Recommendation because of: Bench Warrant n=2 J7B No Recommendation because of: No NYSID n=2 J7C No Recommendation because of: Murder charge n=6 J7D No Recommendation because of: Incomplete interview n=1 Constant model chi-square=25.590

1.465

.0000

4.327

2.361

.0969

10.600

2.361

.0969

10.600

-4.841

.7463

.008

-4.841

.8949

.008

-2.361 .0001

Dependent variable = Failure to Appear 0 = no 1 = yes The reference category is “recommended.” The FTA rate for each of the other categories was compared to the FTA rate for recommended JOs to produce a coefficient, significance level, and odds ratio. These statistics are not calculated for the reference category because it cannot be compared to itself. This statistical technique is called “contrast coding.” A positive logit coefficient and an odds ratio greater than 1.0 indicate that the recommendation category is associated with a higher FTA rate than the reference category (“recommended”). A negative logit coefficient and an odds ratio less than 1.0 indicate that the recommendation category is associated with a lower FTA rate than the reference category. A significance level less than .05 is considered statistically significant.

1 2

FTA was counted up to the sentencing appearance (or to final disposition, if no conviction). 45 cases were excluded from the analysis because of missing data. - 13 -


Turn static files into dynamic content formats.

Create a flipbook
Issuu converts static files into: digital portfolios, online yearbooks, online catalogs, digital photo albums and more. Sign up and create your flipbook.