臺大管理論叢 NTU Management Review VOL.29 NO.1

35 NTU Management Review Vol. 29 No. 1 Apr. 2019 4.2 Data Collection The data was conducted on grain suppliers using a face-to-face questionnaire survey. We assigned this survey to 1284 grain suppliers in Taiwan which were recognized by the Council of Agriculture (COA). Through phone calls and mails, we established each supplier's willingness to take part in the questionnaire survey. In the end, 226 firms completed the questionnaire. The sample profiles are as shown in Table 1. Table 1 Profiles of Participating Members of Grain Supply Chain Demographic profile Number of suppliers Percentage Classified by role Grain distributor 75 33.2 Grain retailer 78 34.5 Grain agent 73 32.3 Total sales revenue (US$) < 0.5 million 69 30.5 0.5 million-5 million 86 38.1 above 5 million 71 31.4 Years of establishment < 5 yrs. 85 37.6 5 yrs. - 10 yrs. 72 31.9 Over 10 yrs. 69 30.5 Position of respondent Top managers 57 25.2 Section managers 82 36.3 Section representatives 87 38.5 4.3 Preliminary Test Structural equation modeling (SEM) involves a measurement model and a structural model (Anderson and Gerbing, 1988). Confirmatory factor analysis (or a measurement), CFA, is one partial analysis of SEM, which defined the link between the constructs and the observed variables. In this study, we use conducted confirmatory factor analysis to assess construct validity of the constructs shown in Table 2. To evaluate the construct validity, we investigated two factors: convergent validity and discriminant validity. For the assessment of the convergent validity, the average variance extracted (AVE) should be less than all the composite reliability (CR) coefficients (Hair Jr, Black, Babin, and Anderson, 2010). In regards to appraising the discriminant validity, the AVE estimated by any pair of constructs should be greater than 0.5. Confirmatory factor analysis (or a measurement)