28.11.2015 Views

ascilite2015

ascilite-2015-proceedings

ascilite-2015-proceedings

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

The questionnaire was pilot tested with a group of students and reviewed by a panel of lecturers for<br />

face and content validity. It comprises 15 statements on perceived ease of use (3 items), perceived<br />

usefulness (3 items), interaction with peers (3 items), lecturer support (3 items) and perceived course<br />

outcomes (3 items). Participants were asked to give their responses to each of the statement on a 5-<br />

point Likert scale, ranging from 1 (strongly disagree) to 5 (strongly agree). When answering the<br />

questions in the questionnaire, the respondents were asked to relate their experience using the LMS<br />

for the e-learning lessons which they had completed. Demographic data such as gender and age<br />

were also collected in the questionnaire.<br />

Statistical Analysis<br />

The analysis of the study was carried out in two stages using a measurement model and structural<br />

model (Anderson & Gerbing, 1998). The first stage involved building a measurement model based on<br />

a confirmatory factor analysis (CFA), and examining the descriptive statistics, and assessing the<br />

validity and reliability. The second stage involved building a structural equation model of the latent<br />

constructs, and testing the hypothesised relationships among the constructs.<br />

Results<br />

Descriptive Statistics<br />

The mean ratings of all the five constructs were between 3.54 and 4.16, and above the mid-point of<br />

3.00 of the scale (see Table 1). This indicated an overall favourable response to the constructs<br />

measured in the study. The standard deviations ranged from .09 to 1.17, which revealed a wide<br />

spread around the mean. The skewness ranged from - .69 to - .05 and kurtosis ranged from - .40 to<br />

.65 were all within Kline’s (2005) suggested cut-offs of absolute values greater than 3 and 10<br />

respectively, indicating univariate normality.<br />

The Mardia’s coefficient in this study was found to be 91.95, below the recommended value of 255<br />

(p(p+2) = 15(17) = 255 where p is the number of observed variables in the study) by Raykov and<br />

Marcoulides (2012). Hence, multivariate normality is met. Therefore, the data is suitable for the<br />

purpose of structural equation modeling.<br />

Table 1: Descriptive statistics of the constructs<br />

Construct Item Mean SD Skewness Kurtosis<br />

Perceived Ease of Use (PE) 3 4.16 1.07 - .45 - .27<br />

Perceived Usefulness (PU) 3 3.81 1.14 - .50 - .08<br />

Lecturer Support (LS) 3 4.61 .97 - .69 .65<br />

Interaction with Peers (IP) 3 3.54 1.17 - .05 - .40<br />

Perceived Course Outcomes<br />

(CO)<br />

3 4.04 1.06 - .69 .32<br />

Exploratory Factor Analysis<br />

The items were subjected to the principle component factor (PCF) analysis with an oblique (promax)<br />

rotation. The Kaiser-Mayer-Olkin (KMO) measure of sampling adequacy was found to be .91,<br />

exceeding the recommended threshold for factor analysis of .6 (Tabachnik & Fidell, 2012). Results<br />

from the Barlett’s test of sphericity provided further support for performing the EFA: Chi-square,<br />

χ 2 (105) = 3147.76, p < .001. The number of resultant five factors was extracted, in line with the<br />

specific variables intended to be measured in the proposed research model. The total variance<br />

explained by the five factors is 84.06%. All the items had standardised factor loadings of over .60, and<br />

the present study accepted this threshold as practical significant (Hair, Black, Babin, Anderson, &<br />

Tatham, 2006).<br />

159<br />

FP:147

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!