Participants
The study's target participants were all Wolaita So University graduating seniors (i.e., third-year students in the colleges of agriculture, natural science, social science and humanities, business and economics, and education and behavioral sciences; fifth-year students in the colleges of engineering and technology, law, and informatics; and sixth-year students in the colleges of health sciences and veterinary medicine). The reason for choosing students in their senior year as the study's target group was that they had already spent a considerable amount of time in college and the dependent variable was cumulative grade point average (CGPA). The numbers of graduating year students from selected university are 1626. The total numbers of selected university students are 1626 where 999 are males and 626 females.
Sample Size And Sampling Technique
First, by using purposive sampling technique, Wolaita Sodo University graduating year students were selected. Following this, they were further divided into groups based on colleges/schools and sexual category for the reasons that there are not an equal number of students in each college/school and that there are significantly less female students than male students in each colleges and schools. From a total of 1626 students, the researchers selected 313 graduating year students as the study's sample using Morgan and Krejcie's 1970 sampling determination table. We as a researcher choose a sample from a population based on a clear approximation by using the sampling determination table created by Morgan and Krejcie. A lot of researchers have used the method for their study and confirmed its validity. Lastly, because randomization is successful in generating correspondent descriptive groups that are basically the same on all applicable variables, simple random sampling processes and a technique of lottery system are also used to select contributors from each colleges and schools in order to avoid unfairness and offer equivalent chance for entire graduation class students (Amin, 2005).
Table 1
College/School
|
Population
|
Sample
|
Male
|
Female
|
Agriculture
|
189
|
36
|
19
|
17
|
Engineering
|
121
|
23
|
18
|
5
|
Natural Computational science
|
166
|
32
|
24
|
8
|
Social Science & Humanities
|
225
|
43
|
25
|
18
|
Business and Economics college
|
393
|
76
|
43
|
33
|
Health Science college
|
275
|
53
|
32
|
21
|
Education & Behavioral science
|
101
|
20
|
10
|
10
|
School of informatics
|
61
|
12
|
6
|
6
|
School of law
|
51
|
10
|
8
|
2
|
School of Veterinary
|
44
|
8
|
7
|
1
|
Total
|
1626
|
313
|
192
|
121
|
Source Wolaita sodo University Registrar office (2021/22)
Procedures
Students may experience significant difficulty speaking English because it is a foreign language to Ethiopian university students. Due to this, two senior language instructors who teach at Wolaita Sodo University in Ethiopia language departments who are native Amharic speakers translated the questionnaire into the nation's official language. The survey measures were translated and reverse-translated from one language to another in accordance with Brislin's (1986) guidelines to guarantee conceptual parity between the original instruments (in English) and their equivalents in Amharic. The work received ethical approval from the research ethics committee of Shaanxi Normal University. After being made aware of the significance and potential applications of the study, the participants’ Ethiopian university (Wolaita Soddo University) gave permission for it to be carried out.
The questionnaire made it very clear to participants that their responses would always remain anonymous. They were also being told they could freely and voluntarily offer their voluntary, informed consent and that they have the opportunity to withdraw or decline any moment to take part in the study. The contributors were prearranged guidelines on the best way to finish the questionnaire and were advised that further data will be gathered later, making it essential for them to submit their student identity card number on the form. The student Cumulative Grade-Point-Averages (GPAs) for the final year of the 2021/22 academic year were collected from the participant university in Ethiopia based on the student identity card number provided on the self-report questionnaires after the questionnaire collection process was complete.
Measures
Three sets of questions made up the self-report questionnaire. The first set on the background characteristics of the students (i.e., demographic variables). The scales Student Teacher Interaction (STI) and Academic Self-Efficacy Scale made up the final two sets of items (ASES). Since the scale was created in a different culture, factor analyses have been undertaken for the student teacher interaction and academic self-efficacy surveys.
Factor Analysis Of Student Teacher Interaction Scale
Confirmatory factor analysis is a method for determining if an item fits the underlying model or not (Lania, 2013). The use of CFA makes it possible to analyze and test theoretical models that describe relationships between manifest variables (Fairchild et al., 2005).
The main component analysis was performed to determine the components on which the items loaded and to name those factors, as well as to evaluate the construction validity. The data's suitability for the analysis was assessed using KMO and the Bartlett's Test of Sphericity (BST). The results' KMO score was 0.821. Factor analysis may be carried out, in accordance with (Hair et al., 2014), when the KMO value was greater than 0.6. The significance of the Chi-squared statistics obtained at the conclusion of the BST showed that the data with many variables had a normal distribution. The significance of the BST was demonstrated (Chi-Square = 1486.603; p = 0.000). These outcomes proved that the STI was a suitable subject for factor analysis. There were most likely four components (conflict, closeness, share feeling and asking for help) that best explained the data. Initial EFA results for 28 items with Eigenvalues showed that four higher-than-1-factor structures could be retrieved and accounted for 53.309% of the total variance. Ten items (8, 9, 10, 11, 12, 13, 18, 19, 21 and 22) were not included since their loadings were insufficient and they were loaded on many factors (Pianta, 2001). Confirmatory factor analysis was conducted and model fit was determined.
All elements in the measurement model for each indication had reasonably high loadings, according to the factor loadings. All of the products had standardized loadings that were higher than 0.50. At a significance threshold of p < 0.01, every factor loading was highly significant. Table (2) and Figure (1) showed the measurement model and fit indices.
Table 2
Fit indices of the CFA proposed four-factor model
Model fit
|
Criteria
|
Measurement model
|
(x2)
|
-
|
234.743
|
DF
|
-
|
129
|
X2/DF
|
≤ 5
|
1.820
|
CFI
|
> .900
|
.923
|
GFI
|
> .900
|
.926
|
AGFI
|
> .900
|
.903
|
TLI
|
> .900
|
.908
|
NFI
|
> .900
|
.901
|
RMSEA
|
≤ .080
|
.051
|
Note: X2 stands for Chi-square, DF for degree of freedom, CFI for comparative fit, GFI for general fit, AGFI for adjusted goodness-of-fit, TLI for Tucker-Lewis fit, NFI for standardized fit, and RMSEA for root-mean-square approximation error. |
Measurement Of Academic Self-efficacy
The Measurement Inventory Scale for Academic self-efficacy, which consists of 40 items and was developed by (Abdal and Muhammad, 2006), was used to measure it. In other words, the scale was modified by using all 40 questions from the Abdal and Muhammad academic self-efficacy scale (2006).
The results of the BST's obtained from Chi-squared tests, which were significant, indicated that the data with multiple variables had a normal distribution. It was determined that the BST was significant (Chi-Square = 2140.956; =0.000). There are most likely five factors (confidence, dependency, recalling, asking for help and not understanding concepts) that can explain the data Five higher-than-1-factor structures could be extracted from the initial EFA results for 40 items having Eigenvalues, and they were responsible for 56.276% of the overall variation.. Due to its insufficient loadings and the fact that they were loaded on numerous criteria, 18 items (1, 4, 8, 10, 11, 12, 13, 14, 16, 20, 21, 22, 24, 25, 26, 32, 33, and 37) was excluded (Abdal and Muhammad, 2006).
All elements in the measurement model for each indication had reasonably high loadings, according to the factor loadings. All of the products had standardized loadings that were higher than 0.50. At a p < 0.01 threshold of significance, all factor loadings were very significant. Table (3) and Figure (2) displayed the measurement model and fit indices
Table 3
Fit indices of the CFA proposed five-factor model
Classic fit
|
Standards
|
Measurement model
|
(x2)
|
-
|
353.927
|
DF
|
-
|
199
|
X2/DF
|
≤ 50
|
1.779
|
CFI
|
> .900
|
.921
|
TLI
|
> .900
|
.909
|
AGFI
|
> .900
|
.901
|
GFI
|
> .900
|
.914
|
NFI
|
> .900
|
.897
|
RMSEA
|
≤ .080
|
.050
|
Note: X2 stands for Chi-square, DF for degree of freedom, CFI for comparative fit, GFI for general fit, AGFI for adjusted goodness-of-fit, TLI for Tucker-Lewis fit, NFI for standardized fit, and RMSEA for root-mean-square approximation error. |
Academic Achievement (Aa)
By using the identity card number of student that they had provided on their self-reporting surveys the Registrar Office of participant University was able to access the official records and determine the total 3, 5 and 6 years average results (CGPAs). The accumulative total average results (CGPAs), which range from 0.00 to 4.00 on a 4-point scale, are computed by country standards of higher education organizations.
Data Analysis
A quantitative data analysis system processed the information obtained by the questionnaire in parallel activity flows. The SPSS 25 version for data processing and the Hayes PROCESS script version 3.5 in SPSS were used to analyze the data. The minimal sample requirements for each analysis were satisfied (Tabachnick & Fidell, 2013). First-time descriptive evaluations, including central tendency (mean, median, mode), and standard deviations, along with a calculation of the data's regularity, were first applied. In order to examine the relationships between and among all of the study variables, Pearson-product moment correlations were established. A T-test was performed to evaluate group differences and determine whether sex was related to the variables employed in the current investigation (i.e., between males and females).The T-test results showed no statistically major mean differences among male and female academic achievement, and the factors were not statistically related to them.
The other hypotheses were assessed via Hayes PROCESS macro version 3.5 model analysis with bias-corrected, accelerated, and centered 95% confidence intervals based on 5,000 bootstrapped samples. The Sobel Test and other mediation analytic methods are not as effective as bootstrapped methods (Hayes, 2019). The bootstrapped Confidence Intervals were produced using the PROCESS script version 3.5 for single mediators with the intention of evaluating the mediation effect. The stated mediation effects were deduced, in Hayes' opinion, in such a way that the mediation effect was deemed significant when the lower and upper 95% CIs were either both above or below zero (i.e., lower and upper CIs did not include a zero), and the outcome was considered non-significant when lower and upper CIs included a zero (Hayes, 2019).