 |
Asia-Pacific Forum on Science Learning and Teaching, Volume 17, Issue 2, Article 9 (Dec., 2016) |
The results of this study are presented in accordance with the research aim indicated above. The findings are presented in two parts, namely: descriptive and inferential findings. The initial findings highlight the science learning outcomes and creative thinking of Grade 11 students before and after a constructionism and neurocognitive-based teaching model, and a conventional teaching model, were used in their educational instruction. This is followed by evaluating the impact of these two teaching models on the Grade 11 students’ science learning outcomes and creative thinking. Finally, the different impacts of the two teaching models are measured.
The descriptive statistics of pre-test vs. post-test of nanotechnology content knowledge, science process skills, scientific attitudes, and creative thinking of both the experimental and control groups for the Grade 11 students are presented in Table 1. All post-test results show an increment compared to the pre-test results after utilizing any of the two teaching models.
Table 1. Descriptive statistics of pre-test vs post-test
Dependent variables
Experimental group (N=45)
Control group (N=34)
Pre-test
Post-test
Pre-test
Post-test
M
SD
M
SD
M
SD
M
SD
NT. content knowledge
11.40
2.43
22.40
2.99
11.59
2.61
16.79
3.22
Sci. process skills
30.18
5.61
33.71
4.16
28.29
3.11
30.35
3.02
Scientific attitudes
92.89
9.79
104.09
8.66
95.15
10.84
96.24
9.65
Creative thinking
174.24
39.42
220.38
66.55
150.41
42.72
163.50
58.24
A 2x2 repeated MANOVA was utilized to analyse the effect of the two teaching models on all of the four dependent variables. The Box’s M was significant implied that the covariance matrices of the dependent variables were not equal across the groups. Therefore the Type 1 error should be considered. The results revealed that there was a significant multivariate effect for interaction between teaching models and time. Pillai’s trace value = .531, F(4, 74) = 20.982, p= .000; partial η2 = .531 showed that this interaction could explain 53.1 percent of variance in the dependent variables. Power to detect the effect was 1.000, which showed that the sample size was adequate. However, the Levene’s Test of Equality of Error Variances showed the difference of error variance across groups on pre-test science process skills, F(1, 77) = 13.389, p= .000, and violated the assumption.
On this line of reasoning, researchers used the nonparametric method for testing this variable. The Mann-Whitney U test showed that before the intervention, science process skills were significantly different between groups (U = 560.000, p = .420). The analysis design was then changed to see whether or not these two groups were equal, and in which particular variables, before intervention, and used those variables as the covariates. A one-way MANOVA for the other three remaining pre-test variables showed that there was no significant difference between groups, F(3, 75) = 2.720, p = .050, with Box’s M = 3.931, p = .709, and all the Levene’s Test of Equality of these three variables were not significant. Therefore, the covariance variable in the post-test analysis was only the pre-test of the science process skills score.
A one way MANOVA for the four dependent variables after intervention (Box’s M = 8.881, p = .503) revealed a significant multivariate main effect for teaching models, Wilks’ λ =.423, F(4, 74) = 25.207, p = .000; the teaching model could explain 57.7 percent of variance in the dependent variables, partial η2 = .577; power to detect the effect was 1.000. Given the significance of the overall test, the univariate main effects were examined. Significant univariate main effects for teaching models were obtained for all dependent variables as shown in Table 2 below.
Since the pre-test of science process skills differed between groups, a MANCOVA design was applied to make sure that post-test differences truly resulted from the treatment, and were not from some other left-over effect of pre-test differences between the groups. The results revealed a significant multivariate main effect for teaching models, Wilks’ λ =.440, F(4, 73) = 23.243, p = .000; the ability to explain variance in dependent variables showed a small decrease, partial η2 = .560; power to detect the effect was 1.000. All the assumptions were met. Significant univariate main effects for teaching models were obtained for all dependent variables as indicated in Table 2 below.
Table 2. Summary of univariate tests results for the two models
Variables
No covariate
With covariate
Nanotechnology content knowledge
F(1,77) = 63.643
p = .000
ηp2 = .453F(1,76) = 57.528
p = .000
ηp2 = .431Science process skills
F(1,77) = 16.256
p = .000
ηp2 = .174F(1,77) = 12.917
p = .000
ηp2 = .145Scientific attitudes
F(1,77) = 14.440
p = .000
ηp2 = .158F(1,77) = 14.319
p = .000
ηp2 = .159Creative thinking
F(1,77) = 15.727
p = .000
ηp2 = .170F(1,77) = 16.320
p = .000
ηp2 = .177
