3.2. Data collection instruments

We adopted 33-items questionnaire originally formulated by Schutte et al. (1998) on the basis of Salovey and Mayer (1990)'s theoretical model of EI. The questionnaire is a one-dimensional construct. A five-point Likert-type response scale ranging from strongly disagree (1) to strongly agree (5) was used to measure emotional intelligence behavior (Zhoc and Chen, 2016). Sample item of EIS is ‘I know when experience my emotions’. The scale has a sound psychometric properties with 0.90 alpha value (Schutte et al., 1998). Subsequent efforts have been made to revalidate the factor solution of the construct and the analysis revealed its multifaceted nature with significance variability in the number of dimensions (Chan, 2008; Zhoc and Chen, 2016). However, two sub-dimensions of EIS, namely, Interpersonal EI and Intrapersonal EI, were examined in the current study. The alpha value of the two EI sub-dimensions are Interpersonal EI, α = .67, Intrapersonal EI, α = .60 and they are adequate for exploratory factor analysis (Hair et al., 2010). The composite reliability (CR) scores of the two sub-constructs ranged from .68 to .72 and exceeded the .60 desired standard, confirming high internal consistency.

We adopted 10-items SDLS formulated by Lounsbury and Gibson (2006) from personality perspective model (Brockett, 1983). The questionnaire is a one-dimensional construct. A five-point Likert-type response scale ranging from strongly disagree (1) to strongly agree (5) was used to measure student's self-directed learning behavior (Lounsbury et al., 2009). The reliability coefficients of the scale ranges from 0.84 to 0.87 when used for high school and college students (Lounsbury et al., 2009). In the current research, the alpha value of the self-directed learning scale exceeded 0.70 required standard (α = .77). The composite reliability (CR) value of the construct is .77 and exceeded the minimum cut-off of .60, reflecting high internal consistency.

The background variables captured in the current study are student demographic variables (age and gender). The academic performance of the students were determined by the achievement scores students obtained at the end of the academic session and it ranges from 0 to 400. Each student was expected to offer four science courses and each course is allocated a maximum score of 100. We used the academic records of the students and the achievement scores were culled from the official result records of these students. The scores of these students were tallied with their responses in the questionnaire using their registration number. The age and gender were retrieved from the bio-information the students supplied when they were completing the questionnaires during data collection.

The present study considered several data screening-related issues. Missing data and outliers were identified in each sub-construct through boxplot (Kwak and Kim, 2017). Univariate normality test for each item was done using skewness and kurtosis with a range of +1.90 to -1.90 at the significance level of 0.05 (Hair et al., 2010). Multivariate normality assumption was check by considering the homoscedasticity issues. In addition, multicollinearity problem was tested using tolerance and variance inflation factor (VIF)’ (Pallant, 2020).

Exploratory Factor Analysis (EFA) were implemented using SPSS 26 to revalidate the factor structures of the EIS and SDLS in Nigerian context (Lorenzo-Seva and Ferrando, 2006). Subsequently, confirmatory factor analysis were employed to determine the measurement model of EIS and SDLS using AMOS 24.0 (Byrne, 2010). Apart from chi-square test (χ²) and the ration of Chi-square to degrees of freedom (χ²/df) which have the limitation of rejecting a model with a huge sample size, other fit indices we employed to gauge the fitness of the model include, Comparative Fit Index (CFI), Goodness of Fit Index (GFI) and Root Mean Square Error of Approximation (RMSEA) (Awang, 2012). According to (Awang, 2012), the cut-off point for the various fit indices included the following; (χ², p > 0.05), (χ²/df < 5.00), (CFI >0.90), (GFI >0.90) and (RMSEA <0.08). Reliability of the scales were determined using Cronbach's alpha coefficient and composite reliability. Due to exploratory factor analysis, alpha value ranges from .60 to .70 is acceptable (Hair et al., 2010), and composite reliability value of .60 and above is satisfactory (Awang, 2012). The unique contributions of predictor variables on the outcome variable and the incremental validity were examined through hierarchal regression analysis.