Navigating the Corridors of Recidivism: A Comprehensive Data Analysis Approach in Formerly Incarcerated Individuals

Problem Description:

The dissertation focuses on exploring the relationship between emotional intelligence and recidivism in individuals with a criminal/incarceration past. The researcher aims to answer specific research questions through a quantitative correlational study with a cross-sectional approach. The study also delves into the rationale behind choosing this research design and explores alternative non-experimental designs.

Solution Presentation:

Research Design and Rationale:

The chosen research design for this dissertation is a quantitative correlational study with a cross-sectional nature. This approach is suitable for investigating predictive relationships among variables. While acknowledging the limitations, such as the inability to prove causation, the correlational design is justified due to its effectiveness in utilizing real-world data. The cross-sectional approach is employed to test relationships within a specific timeframe, aligning with the use of secondary data. Despite limitations like limited data access, the cross-sectional design is preferred for its ability to provide detailed information.

Alternative Designs Considered:

Causal-comparative and descriptive studies were considered but deemed inappropriate. Causal-comparative studies lack suitability due to the requirement for a binary or categorical independent variable. Descriptive studies, though offering a high-quality appraisal, are unsuitable for this study's inferential statistical requirements.

Methodology Section:

The sampling strategy is carefully chosen based on specific inclusion criteria. The use of cross-sectional data analysis is highlighted for its advantages in providing detailed information even when the entire dataset is inaccessible. Probabilistic and random sampling methods were considered but discarded due to difficulties in accessing the target population. A certain sample was obtained through random selection, reflecting the reality of the study.

Sample Size Determination:

G*Power analysis was employed for sample size determination, considering a significance level of 0.05, statistical power of 80%, and a medium effect size. The minimum necessary sample size for Research Question 1 is 98. For Research Question 2, minimum sample size calculation involves considerations of logistic regression, yielding a requirement of at least 120 participants. The study will proceed with the larger value, ensuring a sample size of at least 98 participants.

Instrumentation:

Emotional intelligence is measured using the Schutte Self Report Emotional Intelligence Test (SSEIT). The SSEIT, known for its internal consistency (Cronbach’s alpha of 0.90), utilizes Likert scale responses. The total score is considered for individuals with a criminal/incarceration past. The instrument's reliability is emphasized, with references to established standards for strong reliability.

By presenting the solution in a structured manner, the content becomes more digestible and aligned with academic writing standards.

Data Analysis Plan:

Data Cleaning:

Upon reaching the calculated sample size, the data will be downloaded from Qualtrics to an SPSS file. The cleaning process involves deleting cases where inclusion questions were answered negatively or data is incomplete. Variable values will be verified against Table 4, and the 'number of times arrested' variable will be recoded into a binary variable (0=arrested once/1=arrested more than once) for RQ2.

Descriptive Statistics:

Descriptive statistics will be reported for independent and dependent variables, including gender, ethnicity, education level, binary categories of 'number of times arrested,' and SSEIT items. Measures such as means, ranges, and standard deviations will be presented for 'number of times arrested,' 'number of times in jail/prison,' and overall SSEIT score (Field, 2013).

Research Question 1: Multiple Linear Regression:

RQ1 aims to understand the relationship between demographic factors, arrest history, jail time, and emotional intelligence in formerly incarcerated individuals.

• H01: No statistically significant relationship.
• HA1: Statistically significant relationship.

Assumption Testing for Multiple Linear Regression:

1. Continuous Dependent Variable.
2. Two or More Continuous or Categorical Independent Variables.
3. Independence of Observations.
4. Linear Relationships Between Each IV and the DV.
5. Homoscedasticity (Brusch-Pagan Test) and Normality (Shapiro–Wilk test).
6. Lack of Multicollinearity (VIF test).
7. No Significant Outliers.
8. Residuals are Approximately Normally Distributed.

Research Question 2: Multiple Logistic Regression:

RQ2 investigates the predictive relationship between demographic factors, emotional intelligence, and rearrest status in formerly incarcerated individuals.

• H02: No statistically significant predictive relationship.
• HA2: Statistically significant predictive relationship.

Assumption Testing for Multiple Logistic Regression:

1. Binary Dependent Variable.
2. Linearity between Independent Variables and Log Odds (Box-Tidwell procedure).
3. Absence of Multicollinearity (VIF test).
4. Standardized Residuals for Outliers (Field, 2013).

Multiple Logistic Regression Analysis:

If assumptions are met, the Ordinary Least Squares (OLS) method will be employed for estimation, ensuring consistent and effective results with linear data. In case the minimum sample size of 120 is not reached, the nonparametric bootstrap method will be utilized for robust statistical estimation (Mooney & Duval, 1993).

By outlining the data analysis plan in a structured manner, it becomes easier to follow the logical flow of the research process, from data cleaning to hypothesis testing. This clarity enhances the transparency and reproducibility of the study.