Learning Gain
This section investigates the impact of each of the contributing factors to the test scores and/or learning gain.
A multiple regression model shows the influence each contributing factor makes to the scores. This is useful because it identifies which factors are significant and how it quantifies the degree each factor contributes to the score.
A test score is the absolute score recorded for a test. The learning gain is described as the difference between two successive test scores, therefore learning gains may occur between:
1. pre-test and final post-test,
2. pre-test and interim test and
3. Interim test and final post-test.
The hypothesis focuses upon the constructivist learning techniques, therefore the model concentrates on constructivist tasks. The constructivist tasks are map-based exercises on the touch table using model objects. The contributing factors are conditions that may influence the outcome of these exercises.
These conditions include:
1. Time taken to complete exercise tasks,
2. Number of times each interface feature is used,
3. Prior knowledge and
However, we already know both Prior Knowledge (Section 6.9) and number of group members (Section 6.10) do not significantly affect test scores. Therefore the model will only use: Time taken to complete exercise tasks and Number of times each interface feature is used.
6.11.1
Multiple Regression Model Assumptions
All multiple regression relies on the following assumptions (Quinn & Keough 2002): Outliers removed, collinearity of data, independent errors, random normal distribution of errors, homoscedasticity & linearity of data and non-zero variances.
A review of the standard residuals of the Frequency of use of interface objects and Time
spent on constructivist tasks showed that data contained no outliers (Std. Residual Min=- 1.70, Std. Residual Max = 2.79).
Collinearity statistics indicated that multicollinearity was not a concern (Frequency of
use of interface objects, Tolerance = 1.0, VIF = 1.0; Time spent on constructivist tasks, Tolerance = 1.0, VIF = 1.0). Tolerances approaching 0 indicate a multicollinearity problem, while VIF states the factor by which the sample size must be increased to eliminate multicollinearity (Quinn & Keough 2002).
The data met the assumption of independent errors (Durbin-Watson value = 2.315). A histogram of standardised residuals indicated that the data contained approximately normally distributed errors, as did the normal P-P plot of standardised residuals, which showed points that were not completely on the line, but close. The scatterplot of standardised predicted values showed that the data met the assumptions of homogeneity of variance and linearity.
The data also met the assumption of non-zero variances (Frequency of use of interface
objects, Variance = 207; Time spent on constructivist tasks, Variance = 20.5; interim test score, Variance = 12.4; interim learning gain, Variance=11.61).
6.11.2
Model Results
The regression model uses all 32 participants from the Interactive TableStatic Map
sequence for the interim test and uses all 32 participants from the Static Map
Interactive Table sequence for the Post-test evaluation. A summary of the multiple regression analysis shows Frequency of use of interface objects and Time spent on constructivist tasks directly influence both the interim test score and interim learning gain (Table 19).
Table 19: Multiple Regression Analysis investigating relationship between both Time spent on constructivist tasks and Frequency of use of interface objects to learning gain and test score. A significant relationship exists for the interim test for both the test score and learning gain (p<0.01)
Test Score Learning Gain Interim Test F(2, 29) = 17.5, R2=0.547, RAdjusted=0.4907, p < .001 F(2,29)=5.90, R2=0.289, R2Adjusted=0.240, p<0.01 Post-test F(2,29)=2.458, R2=0.1449, R2Adjusted=0.08597, p=0.1033 n.s. F(2,29)=3.133, R2=0.1777, R2Adjusted=0.1209, p=0.0586 n.s
Both Frequency of use of interface objects and Time spent on constructivist tasks explain a significant amount of the interim test score (F(2, 29) = 17.5, p < .001, R2=0.547, RAdjusted=0.04907).
The analysis shows that Frequency of use of interface objects significantly predicts the
value of interim test score (Beta = 0.357, t(29) = 2.86, p<0.01), likewise Time spent on
constructivist tasks significantly predicts the value of interim test score (Beta = 0.654 t(29) = 5.23, p < .001).
Frequency of use of interface objects and Time spent on constructivist tasks explain 54.7% of the variance of the interim test score. The Frequency of use of interface objects explains
12.35% while the Time spent on constructivist tasks explains 42.36%
Based upon the multiple regression analysis, the interim test score is estimated by: Interim Test Score = 2.658 + 0.087 Frequency of use of interface objects
Both Frequency of use of interface objects and Time spent on constructivist tasks explain a significant amount of the Interim Learning gain (F(2,29)= 5.90, R2=0.289, R2Adjusted=0.240, p<0.01).
The analysis shows that Frequency of use of interface objects significantly predicts the
value of interim learning gain (Beta=0.605, t(29)=2.37,p<0.01), however Time spent on
constructivist tasks does not significantly predict the value of interim learning gain (Beta=-0.088,t(29)=-0.346,p=0.08n.s.).
Frequency of use of interface objects and Time spent on constructivist tasks explain 28.9% of the variance of the interim learning gain, all explained by the Frequency of use of interface objects.
Based upon the multiple regression analysis, the interim learning gainis estimated by:
interim learning gain = 1.437 + 0.59 Frequency of use of interface objects -0.088 Time spent on constructivist tasks
Significant scores for interim test score and interim learning gain are expected because Time spent on constructivist tasks is significant only for interim results (Section 6.8.4).