Individual Confidence Intervals and Accuracy of Risk Assessment Evaluations

Statistical Technique: Confidence Interval

Individual Confidence Intervals and Accuracy of Risk Assessment Evaluations

Purpose of Research

The psychological assessments undertaken to evaluate the risk of recidivism can have severe outcomes for the patients assessed and potential victims. Hence, it is necessary that assessors specify the accuracy and error made in such evaluations. Few commentators have raised concerns over the idea of applying recidivism rates to individual cases projected from group data given that individual confidence intervals are large. The purpose of this article is to prove the statement that individual confidence intervals offer modest information about the accuracy of a risk evaluation. The confidence intervals for recidivism forecast will almost vary from 0 to 1 when the result is dichotomous. As a result, other indicators of predictive accuracy are desirable, some of which are qualitative (Hanson & Howard, 2010).

Type and Methodology of Research Variables

The confidence-interval of 95% is the gap within which 95% of the likely results are likely to be found. The result where the individual is either not reoffended or reoffends is an instance of the Bernoulli process. The only bound of the Bernoulli distribution is p, the probability of recidivism or, “positive” outcome. Therefore, a binary variable can only take value 1 (reoffends) or 0 (not reoffend). The confidence interval of 95% is (0, 1) around p, whenever p lies between .95 and .05. The interval is (0, 0) whenever p is .05 or lower. When interval is (1, 1) the p is .95 or higher. There is no need to empirically prove the fact as it is true by definition (Hanson & Howard, 2010).

Statistical Technique Utilized

Confidence Intervals - In the interpretation of a clinical trial, statistical significance is an important element that ensures that the result has a good chance of being real and not be a coincidence (Babchishin & Hanson, 2009).

Reliability & Validity

Hart et.al (2007) has pointed out several limitations; however, Cooke & Michie (2009) have provided a correct analysis of using confidence interval. It is valid that confidence interval ranges (1, 1) when applied to an individual during actuarial assessments. This phenomenon is not limited to actuarial risk evaluation. It is applicable to any predictions for all dichotomous results. Therefore, using confidence intervals is reliable and valid for proving such arguments (Agresti & Caffo, 2000).

Results

According to the findings of the research, the range of confidence intervals for recidivism results, for single case, is between 0-1. This is true when the probability is between .95 and .05 and results are dichotomous. As a result, there is a need of other qualitative indicators in order to determine the error and accuracy of risk assessments (Hanson & Howard, 2010).

Research Recommendations

There are some negligible factors neglected in the research; however, nothing can be said for sure until any new and more comprehensive research is undertaken to provide better outcomes for risk assessments. Recommendations gain credibility when new external factors are examined to proof their plausibility and implication for past ...