5.7.1 Item Generation Results
Only the 3D authenticity construct and its scale have been developed by this research. The remaining scales have been modified based on well known scales that have high reliability and validity. Moreover, each construct is multi-item, as Churchill (1979) recommends. The questionnaire contains 5-point Likert-type scales, anchored by “strongly disagree” and “strongly agree”. To measure the control construct, this thesis developed a five-item scale that centres on users’ ability to rotate and zoom in or out and rotate the virtual model based on Liu’s (2003); McMillan and Hwang’s (2002); and Song and Zinkhan’s (2008) scales. To measure vividness of the animated colours, this thesis developed a four-item scale based on Peg et al’s. (2004), Steure’s (1992); Kim and Biocca’s (1997); and Coyle and Thorson’s (2001) studies. The items map how closely the simulated sensory information reflects the real product. For 3D telepresence, this thesis used a modified version of Kim and Biocca’s (1997) scale with four items. This thesis could not find an existing scale to measure 3D authenticity so it developed a new ten-item scale (based on the literature and focus groups). To measure hedonic and utilitarian values Babin et al.’s (1994) scale and Fiore et
119 used. To measure behavioural intention, a modified version of Fiore and colleagues’ (2005a) scale was used. In total, the first draft of the questionnaire comprised of 66 items. See Table 5.12.
Table 5.12 The constructs and the number of initial items
Construct Number of initial items
3D Control 6 items
3D Animated colours 5 items
3D Telepresence 4 items
3D Authenticity 10 items
3D Hedonic values 11 items
3D Utilitarian values 6 items
3D Behavioural intention 7 items
2D Hedonic values 11 items
2D Utilitarian values 6 items
5.7.2 Purifying Measures
The third step in Churchill’s (1979) paradigm is to purify the measurement scales via checking the reliability and validity for all scale items. McDaniel and Gates (2006, p. 224-227) define validity as “the degree to which what the researcher was trying to measure was actually measured”. This thesis conducted two types of validity, during the preliminary stages and before conducting the main survey; faces validity and content validity. On the other hand, McDaniel and Gates (2006, p. 222) define reliability as “the degree to which measures are free from random error and, therefore, provide a consistent data”. This thesis measured the reliability test by Cronbach’s alpha which should be at least 0.70 or more (Hair et
al., 2006). Using validity and reliability measurements discussed above, the scales have been purified using the following two ways.
120 Qualitative Assessment
During the qualitative assessment, this study assessed two types of validity. First, face validity, “the degree to which a measurement seems to measure what it is supposed to measure” (McDaniel and Gates, 2006, p. 225). Secondly, content validity, “the responsiveness, or sampling adequacy, of the content of the measurement instrument” (McDaniel and Gates, 2006, p. 225) have been conducted for all constructs and especially for 3D authenticity construct by submitting the items to evaluations by academics (lecturers in online retailing and Ph.D. students). These respondents considered the items (questions) relevant and adequate for measuring the constructs.
Quantitative Assessment
In analysing the pilot study (n = 50), this thesis conducted Cronbach alpha reliability and factor analysis for each construct alone. The 3D authenticity construct, Cronbach’s alpha, reveals a value of 0.873, which is above the recommended threshold (0.7, according to Hair et al., 2006). Moreover, the result of item-to-total correlation coefficient alpha shows that only five items correlated above 0.30, and the other five items which failed to achieve the correlation standards of 0.30 were discarded. This study ran an exploratory factor analysis on the remaining five items (n = 50) to check if the 3D authenticity scale is supported by the data. The KMO statistic reveals 0.793, above the minimum recommended value of 0.60 (Kaiser, 1974) and the Barlett’s test of sphericity yielded an χ2 value of 93.752 (p < 0.001), suggesting an acceptable level of the common variance. Cronbach’s alpha of the remaining five items was 0.843, which is above the recommended minimum threshold (0.7). The five retained items each commenced with “after surfing the 3D sites”, and then obtained responses to the following: “3D creates a product experience similar to the one I would have when shopping in a store”; “3D lets me feel like if I am holding a real laptop and rotating it (i.e. virtual affordance)”; “3D lets me feel like I am dealing with a salesman who is responding to my orders”; “3D lets me see the laptop as if it was a real one”; and “Being able to zoom in/out and rotate the laptop let me visualise how the laptop might look in an offline retailer”. The confirmatory factor analysis results are explained in the results chapter.
121 Table 5.13 shows the Kaiser-Mayer-Olkin (KMO) measurement of sampling adequacy and total variance explained. The KMO shows an acceptable level of adequacy (above .60 according to Hair et al., 2006) for all constructs used in this study. Control and animated colours total variance explained were less than the recommended value (0.6 and above). However, all the items of control and animated colours constructs loaded above 0.5, so this study kept them for more analysis (Hair et al., 2006). Hedonic values result revealed that this construct is a multi-dimensional construct (i.e., this construct has items that loaded into two factors). However, in theory, hedonic value is a unidimensional scale. Results in Table 5.14 show that after dropping the items that have communalities below 0.6 hedonic value construct loaded into one factor. The Bartlett Test of Sephericity was significant for all constructs, which indicating correlations presence among all constructs.
Table 5.13 Measurement of sampling adequacy and total variance
construct No. of items Cronbach alpha Sample size No. of factors KMO Variance explained Control 6 .834 50 1 .834 56.101% Animated colours 5 .766 50 1 .677 52.256% 3D Telepresence 4 .843 50 1 .826 62.111% 3D Authenticity 5 .843 50 1 .793 68.501% 3D Hedonic value 11 .917 50 2 .877 66.604% 3D Utilitarian value 6 .879 50 1 .816 63.268 Behavioural intention 7 .911 50 1 .864 66.001% 2D Hedonic values 11 .951 50 2 .897 77.942% 2D utilitarian values 6 .951 50 1 .876 80.937%
122 Table 5.14 Measurement of sampling adequacy and total variance
construct No. of items Cronbach alpha Sample size No. of factors KMO Variance explained Control 6 .834 50 1 .834 56.101% Animated colours 5 .766 50 1 .677 52.256% 3D Telepresence 4 .843 50 1 .826 62.111% 3D Authenticity 5 .843 50 1 .793 68.501% 3D Hedonic value 6 .879 50 1 .833 62.562% 3D Utilitarian value 6 .879 50 1 .816 63.268% Behavioural intention 7 .911 50 1 .864 66.001% 2D Hedonic values 6 .937 50 1 .903 76.889% 2D utilitarian values 6 .951 50 1 .876 80.937%