Analysis and resultsMeasurement mod

Analysis and results

Measurement models

The two-step structural model approach of [4] Anderson and Gerbing (1988) was used. This approach advocates estimating smaller measurement models before the structural model. The scales were divided into two subgroups consisting of exogenous variables (manufacturer brand benefits) and endogenous variables (relationship outcomes of manufacturer brands). A confirmatory factor analysis (CFA) assessed the unidimensionality, reliability, convergent and discriminant validity of these scales which involved the deletion of weak and cross-loading items. The structural modelling used LISREL 8.54 with maximum likelihood estimation and covariance matrix input.

An inspection of the correlations and the item-to-total correlations showed three constructs had high alpha coefficients, but the fourth construct, financial benefits had a low alpha coefficient and low factor loadings. Further examination of this fourth construct revealed a lack of unidimensionality because two of the four items had low item-to-total correlations. As a result, only two of these items were used in the financial benefits scale. Some items in the manufacturer support and customer expectation scales also had low item-to-total correlations and were deleted accordingly.

The manufacturer brand benefit constructs all had high construct validity, as shown by the high item loadings onto the theorised construct, item-to-total correlations and coefficient alpha estimates. The composite reliability and variance extracted estimates also have strong reliability exceeding the 0.6 and 0.5 thresholds respectively ([8] Bagozzi et al. , 1991). High factor loadings and significant t values for the items demonstrate convergent validity. Discriminant validity between the four constructs is evident as the average variance extracted exceeds the correlation squared for each construct pairing except for manufacturer support and financial benefits. A further test of discriminant validity was conducted using an alternative measurement model where the correlation between these two constructs was set to 1. There was a significant change in the chi-squared statistic compared to the four-factor model which confirmed satisfactory discriminant validity. Thus the manufacturer brand benefits model confirms the theoretical direction of the qualitative findings. The fit indices are χ 2 (59)=262, p =0.000, CFI=0.99, NNFI=0.99, RMSEA=0.065 and are within acceptable limits ([33] Hair et al. , 1998).

Estimating the six endogenous constructs measurement model initially showed a poor fit with the data. The poor fit was attributable to weak and cross-loading items associated with the dependence and co-operation constructs. Therefore the model was re-specified by deleting the dependence and co-operation constructs and some items from the satisfaction and performance constructs. The revised four-factor model, consisting of satisfaction, performance, trust and commitment, had a satisfactory fit to the data: χ 2 (48)=213, p =0.000, χ2 /df=4.4, CFI=0.99, NNFI=0.98 and RMSEA=0.065.

The endogenous constructs also have good reliability as shown by the variance extracted estimates and item-to-total correlations. Each item loads onto the correct construct indicating unidimensionality, while the significant t- values high item loadings demonstrate convergent validity. For each construct, the average variance extracted between each pair of constructs exceeds their squared correlations confirming discriminant validity ([23] Fornell and Larcker, 1981). The findings also confirm that extant supplier satisfaction, performance, trust and commitment measures are relevant to manufacturer brands. Table II [Figure omitted. See Article Image.] summarises the final measurement constructs and items and the Appendix, Table AI provides a correlation matrix.

Hasil (Bahasa Indonesia) 1:

Analisis dan hasilPengukuran modelPendekatan struktural model dua langkah Anderson [4] dan Gerbing (1988) digunakan. Pendekatan ini Advokat memperkirakan model pengukuran yang lebih kecil sebelum model struktural. Timbangan dibagi dalam dua subkumpulan yang terdiri dari eksogen variabel (produsen merek manfaat) dan variabel endogen (hubungan hasil merek produsen). Analisis faktor konfirmasi (CFA) dinilai unidimensionality, keandalan, konvergen dan validitas diskriminan Timbangan ini yang melibatkan penghapusan item yang lemah dan salib-loading. Pemodelan struktural digunakan LISREL 8.54 dengan maksimum kemungkinan estimasi dan kovarians matriks masukan.Pemeriksaan korelasi dan korelasi item-untuk-total menunjukkan konstruksi tiga memiliki tinggi alpha koefisien, tapi keempat membangun, keuntungan keuangan memiliki koefisien alpha rendah dan rendah faktor bongkar muat. Pemeriksaan lebih lanjut membangun keempat ini mengungkapkan kurangnya unidimensionality karena dua item empat rendah item-untuk-total korelasi. Sebagai hasilnya, hanya dua item yang digunakan dalam skala keuntungan keuangan. Beberapa item di produsen dukungan dan pelanggan timbangan harapan juga telah rendah item-untuk-total korelasi dan dihapus sesuai.The manufacturer brand benefit constructs all had high construct validity, as shown by the high item loadings onto the theorised construct, item-to-total correlations and coefficient alpha estimates. The composite reliability and variance extracted estimates also have strong reliability exceeding the 0.6 and 0.5 thresholds respectively ([8] Bagozzi et al. , 1991). High factor loadings and significant t values for the items demonstrate convergent validity. Discriminant validity between the four constructs is evident as the average variance extracted exceeds the correlation squared for each construct pairing except for manufacturer support and financial benefits. A further test of discriminant validity was conducted using an alternative measurement model where the correlation between these two constructs was set to 1. There was a significant change in the chi-squared statistic compared to the four-factor model which confirmed satisfactory discriminant validity. Thus the manufacturer brand benefits model confirms the theoretical direction of the qualitative findings. The fit indices are χ 2 (59)=262, p =0.000, CFI=0.99, NNFI=0.99, RMSEA=0.065 and are within acceptable limits ([33] Hair et al. , 1998).Estimating the six endogenous constructs measurement model initially showed a poor fit with the data. The poor fit was attributable to weak and cross-loading items associated with the dependence and co-operation constructs. Therefore the model was re-specified by deleting the dependence and co-operation constructs and some items from the satisfaction and performance constructs. The revised four-factor model, consisting of satisfaction, performance, trust and commitment, had a satisfactory fit to the data: χ 2 (48)=213, p =0.000, χ2 /df=4.4, CFI=0.99, NNFI=0.98 and RMSEA=0.065.The endogenous constructs also have good reliability as shown by the variance extracted estimates and item-to-total correlations. Each item loads onto the correct construct indicating unidimensionality, while the significant t- values high item loadings demonstrate convergent validity. For each construct, the average variance extracted between each pair of constructs exceeds their squared correlations confirming discriminant validity ([23] Fornell and Larcker, 1981). The findings also confirm that extant supplier satisfaction, performance, trust and commitment measures are relevant to manufacturer brands. Table II [Figure omitted. See Article Image.] summarises the final measurement constructs and items and the Appendix, Table AI provides a correlation matrix.

Hasil (Bahasa Indonesia) 2:

Analisis dan hasil

pengukuran model

Dua-langkah model struktural pendekatan [4] Anderson dan Gerbing (1988) digunakan. Pendekatan ini menganjurkan memperkirakan model pengukuran yang lebih kecil sebelum model struktural. Timbangan dibagi menjadi dua kelompok yang terdiri dari variabel eksogen (manfaat merek produsen) dan variabel endogen (hasil hubungan produsen merek). Sebuah analisis faktor konfirmatori (CFA) menilai unidimensionality, kehandalan, konvergen dan diskriminan validitas skala ini yang melibatkan penghapusan lemah dan lintas-loading barang. Pemodelan struktural digunakan LISREL 8.54 dengan estimasi kemungkinan maksimum dan masukan kovarians matriks.

Sebuah pemeriksaan korelasi dan item-to-Total korelasi yang menunjukkan tiga konstruk memiliki koefisien alpha yang tinggi, tetapi konstruk keempat, keuntungan finansial memiliki koefisien alpha rendah dan rendah beban faktor. Pemeriksaan lebih lanjut dari konstruk keempat ini mengungkapkan kurangnya unidimensionality karena dua dari empat item memiliki korelasi item-to-total yang rendah. Akibatnya, hanya dua item ini digunakan dalam skala keuntungan finansial. Beberapa item dalam mendukung produsen dan harapan pelanggan skala juga memiliki rendah item-to-Total korelasi dan dihapus sesuai.

Manfaat merek produsen membangun semua memiliki validitas konstruk yang tinggi, seperti yang ditunjukkan oleh beban tinggi item ke konstruk teori, item-untuk korelasi -Total dan estimasi koefisien alpha. Komposit keandalan dan varians diekstraksi perkiraan juga memiliki kehandalan yang kuat melebihi 0,6 dan 0,5 ambang masing-masing ([8] Bagozzi et al., 1991). Faktor loadings yang tinggi dan nilai-nilai t yang signifikan untuk item menunjukkan validitas konvergen. Validitas diskriminan antara empat konstruk terbukti sebagai varians rata diekstraksi melebihi korelasi kuadrat untuk setiap pasangan membangun kecuali dukungan produsen dan keuntungan finansial. Tes lebih lanjut dari validitas diskriminan dilakukan dengan menggunakan model pengukuran alternatif mana korelasi antara dua konstruksi ini ditetapkan untuk 1. Ada perubahan yang signifikan dalam statistik chi-squared dibandingkan dengan model empat faktor yang dikonfirmasi validitas diskriminan memuaskan. Sehingga produsen manfaat merek Model menegaskan arah teoritis temuan kualitatif. Indeks fit yang χ 2 (59) = 262, p = 0,000, CFI = 0,99, NNFI = 0.99, RMSEA = 0,065 dan berada dalam batas yang dapat diterima ([33] Hair et al., 1998).

Memperkirakan pengukuran enam konstruksi endogen Model awalnya menunjukkan cocok miskin dengan data. Fit miskin disebabkan lemah dan lintas-loading item yang terkait dengan ketergantungan dan kerjasama konstruksi. Oleh karena itu model itu kembali ditentukan dengan menghapus ketergantungan dan kerjasama konstruksi dan beberapa item dari kepuasan dan kinerja konstruksi. Direvisi model empat faktor, yang terdiri dari kepuasan, kinerja, kepercayaan dan komitmen, memiliki fit memuaskan data: χ 2 (48) = 213, p = 0,000, /df=4.4 χ2, CFI = 0,99, NNFI = 0.98 dan RMSEA = 0,065.

The konstruksi endogen juga memiliki keandalan yang baik seperti yang ditunjukkan oleh varians diekstraksi perkiraan dan korelasi item-to-Total. Setiap beban item ke konstruk benar menunjukkan unidimensionality, sedangkan nilai t signifikan beban tinggi item menunjukkan validitas konvergen. Untuk setiap konstruk, varians rata diekstraksi antara setiap pasangan konstruksi melebihi korelasi kuadrat mereka mengkonfirmasikan validitas diskriminan ([23] Fornell dan Larcker, 1981). Temuan juga mengkonfirmasi bahwa kepuasan pemasok yang masih ada, kinerja, kepercayaan dan komitmen langkah-langkah yang relevan dengan produsen merek. Tabel II [Gambar dihilangkan. Lihat Pasal Gambar.] Merangkum konstruksi pengukuran akhir dan barang-barang dan Lampiran, Tabel AI menyediakan matriks korelasi.