IV. FINDINGS OF THE STUDYAs discuss

IV. FINDINGS OF THE STUDY

As discussed in the study of Garver and Mentzer (1999), the Structural Equation Model (SEM) is a powerful technique that combines the measurement model (confirmatory factor analysis) and the structural model (path analysis) into a simultaneous statistical test and provides a measurement theory and a structural theory (Hair, et al., 2006). In this report, the SEM was used as the main statistical analysis tool to purify the measurement items and AMOS 5.0 was used for testing the hypothesis relationship.

Confirmatory Factor Analysis (CFA)

According to Anderson and Gerbing (1988), and Hair, et al. (2006), the first assessment should be any structural model that exists with an acceptable goodness-of-fit. Thus, it could begin by fitting a CFA model that includes covariance between all pairs of latent factors. In this paper, the overall fit for the base



measurement model were poor – GFI=0.886, AGFI=0.861, TLI=0.873, CFI=0.884, RMSEA=0.057 and CMIN/DF=2.693. In order to improve the model fit indices, items PBR4, PBR7, KNLEG2, KNLEG3, and KNLEG4 were deleted from the base model due to the lower factor loading and high modification indices. After deletion, the overall fit for final the measurement model in the calibration sample was excellent, with GFI=0.928, AGFI=0.913, TLI=0.917, CFI=0.927, RMSEA=0.045 and CMIN/DF=2.693.

Testing the Hypotheses

Once an acceptable measurement model is available, the structural model evaluation should be able to start. The results of the structural model show that the model achieved a good level of fit, χ2 = 902.067, χ 2 / df = 2.693, GFI = 0.928, AGFI = 0.913, TLI = 0.917, CFI = 0.927, RMSEA = 0.045. The result also reported that 13.6 per cent of the variance associated with complaint intention was accounted for by its three predictors: knowledge of consumer rights and consumer agencies, perception on business practices and responsiveness to complaint and importance of product. Accordingly, it was determined that 35.3 percent of the variation in complaint action was accounted for by its four predictors including complaint intention.

Testing the Hypotheses on the Complaint Intention

Examining the relationship between the independent variables and the complaint intention, Hypothesis 1a, (see Table 1) regarding the knowledge of consumer rights and consumer agencies on the complaint intention, was supported (P = 0.000, β = 0.295). It explained that if consumers with more knowledge of consumer rights and consumer agencies, they would be more likely to make complaints. This result approved the suggestion from Day and Landon (1976). The result did not support Hypothesis 2a (see Table 1) about the perception on business practices and responsiveness to complaint (p = 0.370, β = 0.039). This means that consumers’ perception on business practices and responsiveness to complaints has no significant relationship with the complaint intention. This result was consistent with the study of Halstead and Droge (1991). Referring to Hypothesis 3a (see Table 1), the results show that the importance of product significantly influences the complaint intention (p = 0.000; β = 0.219). This result indicates that if consumers perceive that the product is expensive, is used frequently, is seen by others and is used for a long period of time, they are more likely to intend to make a complaint. This result is proved in the study of Day (1977).

Results (Indonesian) 1:

IV. TEMUAN-TEMUAN DARI STUDISeperti telah dibahas dalam studi Garver dan Mentzer (1999), Model persamaan struktural (SEM) adalah teknik yang kuat yang menggabungkan model pengukuran (konfirmasi faktor Analisis) dan struktural model (analisis jalur) ke simultan statistik tes dan menyediakan sebuah teori pengukuran dan teori struktural (rambut, et al., 2006). Dalam laporan ini, SEM digunakan sebagai alat analisis statistik utama untuk memurnikan item pengukuran dan AMOS 5.0 digunakan untuk pengujian hipotesis hubungan.Analisis faktor konfirmasi (CFA)Menurut Anderson dan Gerbing (1988), dan rambut, et al. (2006), penilaian pertama harus model struktural yang ada dengan kebaikan diterima-dari-fit. Dengan demikian, ini bisa mulai dengan pas model CFA yang mencakup kovarians antara semua pasang laten faktor. Dalam tulisan ini, secara keseluruhan cocok untuk dasar model pengukuran yang miskin – GFI = 0.886, AGFI = 0.861, TLI = 0.873, CFI = 0.884, RMSEA = 0.057 dan CMIN DF = 2.693. Untuk meningkatkan indeks model yang cocok, item PBR4, PBR7, KNLEG2, KNLEG3, dan KNLEG4 dihapus dari basis model karena loading faktor lebih rendah dan tinggi modifikasi indeks. Setelah penghapusan, secara keseluruhan cocok untuk akhir model pengukuran dalam sampel kalibrasi yang sangat baik, dengan GFI = 0.928, AGFI = 0.913, TLI = 0.917, CFI = 0.927, RMSEA = 0.045 dan CMIN DF = 2.693.Pengujian hipotesisOnce an acceptable measurement model is available, the structural model evaluation should be able to start. The results of the structural model show that the model achieved a good level of fit, χ2 = 902.067, χ 2 / df = 2.693, GFI = 0.928, AGFI = 0.913, TLI = 0.917, CFI = 0.927, RMSEA = 0.045. The result also reported that 13.6 per cent of the variance associated with complaint intention was accounted for by its three predictors: knowledge of consumer rights and consumer agencies, perception on business practices and responsiveness to complaint and importance of product. Accordingly, it was determined that 35.3 percent of the variation in complaint action was accounted for by its four predictors including complaint intention.Testing the Hypotheses on the Complaint IntentionExamining the relationship between the independent variables and the complaint intention, Hypothesis 1a, (see Table 1) regarding the knowledge of consumer rights and consumer agencies on the complaint intention, was supported (P = 0.000, β = 0.295). It explained that if consumers with more knowledge of consumer rights and consumer agencies, they would be more likely to make complaints. This result approved the suggestion from Day and Landon (1976). The result did not support Hypothesis 2a (see Table 1) about the perception on business practices and responsiveness to complaint (p = 0.370, β = 0.039). This means that consumers’ perception on business practices and responsiveness to complaints has no significant relationship with the complaint intention. This result was consistent with the study of Halstead and Droge (1991). Referring to Hypothesis 3a (see Table 1), the results show that the importance of product significantly influences the complaint intention (p = 0.000; β = 0.219). This result indicates that if consumers perceive that the product is expensive, is used frequently, is seen by others and is used for a long period of time, they are more likely to intend to make a complaint. This result is proved in the study of Day (1977).