The correlation of X, and ~5. a het

The correlation of X, and ~5. a heterotrait-heteromethod correlation, is diagrammed in Figure 2. There are four sources of intercorrela-tion between these two measures. First, that due to traits is a product of the path from TZ to X5, the path from T, to X,, and the intercorrelation between T, and T2. Second, the intercorrelation due to commonality of methods is the product of
the path from2 to X 5, the path from M, to X,, and the intercorrelation between M, and M2. Thirdly, the intercorrelation between X, and Xs due to the intercorrelation of T, and M is the
product of the path from MZ to X s, the path from T, to X,, and the intercorrelation between T, andM2- Similarly, the intercorrelation be(1970). An example of the Werts and Linn pro-posal has been given by Kalleberg and Kluegel (1975), and Kenny (1976) has suggested tests of hypotheses concerning various elements of the path analytic formulation. However, a single full presentation of the purposes which the path analysis approach to MTMM matrices can serve is necessary to facilitate its use by a greater number of researchers interested in the construct validity of their instruments. The objective of the present paper is to provide that presentation.Several problems have been noted with respect to the Campbell-Fiske criteria. First, and perhaps most obvious, is the subjectivity of the criteria employed; all that is possible is a greater than or lesser than comparison of correlations in various portions of the matrix. Second, this subjective criterion becomes almost impossible when more than a relatively few traits or methods are involved. Somewhat more subtle are problems which occur when the variables are measured with different reliabilities. What may appear to be discriminant or convergent validity for some traits and not others may simply be due to differential reliabilities. Finally, the implicit assumption in the Campbell-Fiske formulation that traits and methods are uncorrelated is probably not tenable in most situations; further more, it is difficult, if not impossible, to evaluate this assumption by visual examination of the MTMM matrix (Alwin, 1974; Kalleberg & Kluegel, 1975). The confirmatory factor analysis described and illustrated in the present paper allows a more complete analysis of the extent to
which each variable is affected by trait and method variance.

Hasil (Bahasa Indonesia) 1:

Korelasi dari X, dan ~ 5. korelasi heterotrait-heteromethod, teliti menggambar mesin pada gambar 2. Ada empat sumber intercorrela-tion antara langkah-langkah dua. Pertama, itu karena sifat adalah produk dari jalan dari TZ ke X 5, jalan dari T, x,, dan intercorrelation antara T, dan T2. Kedua, intercorrelation karena kesamaan metode adalah produk darifrom2 jalan ke X 5, jalan dari M, ke X,, dan intercorrelation antara M, dan M2. Ketiga, intercorrelation antara X, dan Xs karena intercorrelation T, dan Mproduct of the path from MZ to X s, the path from T, to X,, and the intercorrelation between T, andM2- Similarly, the intercorrelation be(1970). An example of the Werts and Linn pro-posal has been given by Kalleberg and Kluegel (1975), and Kenny (1976) has suggested tests of hypotheses concerning various elements of the path analytic formulation. However, a single full presentation of the purposes which the path analysis approach to MTMM matrices can serve is necessary to facilitate its use by a greater number of researchers interested in the construct validity of their instruments. The objective of the present paper is to provide that presentation.Several problems have been noted with respect to the Campbell-Fiske criteria. First, and perhaps most obvious, is the subjectivity of the criteria employed; all that is possible is a greater than or lesser than comparison of correlations in various portions of the matrix. Second, this subjective criterion becomes almost impossible when more than a relatively few traits or methods are involved. Somewhat more subtle are problems which occur when the variables are measured with different reliabilities. What may appear to be discriminant or convergent validity for some traits and not others may simply be due to differential reliabilities. Finally, the implicit assumption in the Campbell-Fiske formulation that traits and methods are uncorrelated is probably not tenable in most situations; further more, it is difficult, if not impossible, to evaluate this assumption by visual examination of the MTMM matrix (Alwin, 1974; Kalleberg & Kluegel, 1975). The confirmatory factor analysis described and illustrated in the present paper allows a more complete analysis of the extent towhich each variable is affected by trait and method variance.

Hasil (Bahasa Indonesia) 2:

Korelasi dari X, dan ~ 5. korelasi heterotrait-heteromethod, yang digambarkan pada Gambar 2. Ada empat sumber intercorrela-tion antara kedua ukuran ini. Pertama, bahwa karena sifat-sifat adalah produk dari jalan dari TZ ke X5, jalan dari T, untuk X ,, dan intercorrelation antara T, dan T2. Kedua, intercorrelation karena kesamaan metode adalah produk dari
jalan from2 ke X 5, jalan dari M, untuk X ,, dan intercorrelation antara M, dan M2. Ketiga, intercorrelation antara X, dan Xs karena intercorrelation dari T, dan M adalah
produk dari jalan dari MZ ke X s, jalan dari T, untuk X ,, dan intercorrelation antara T, andM2- Demikian pula, intercorrelation menjadi (1970). Sebuah contoh dari Werts dan Linn pro-cara pembuangan telah diberikan oleh Kalleberg dan Kluegel (1975), dan Kenny (1976) telah menyarankan tes hipotesis mengenai berbagai elemen jalan formulasi analitik. Namun, presentasi penuh tunggal dari tujuan yang jalan pendekatan analisis dengan matriks MTMM dapat melayani diperlukan untuk memudahkan penggunaannya oleh sejumlah besar peneliti tertarik pada validitas konstruk instrumen mereka. Tujuan dari penelitian ini adalah untuk memberikan yang masalah presentation.Several telah dicatat sehubungan dengan kriteria Campbell-Fiske. Pertama, dan mungkin yang paling jelas, adalah subjektivitas kriteria yang digunakan; semua yang mungkin adalah lebih besar dari atau lebih rendah daripada perbandingan korelasi di berbagai bagian dari matriks. Kedua, kriteria subjektif ini menjadi hampir mustahil ketika lebih dari relatif sedikit ciri-ciri atau metode yang terlibat. Agak lebih halus adalah masalah yang terjadi ketika variabel diukur dengan reliabilitas yang berbeda. Apa yang mungkin tampak diskriminan atau konvergen validitas untuk beberapa sifat dan tidak yang lain mungkin hanya karena untuk diferensial reliabilitas. Akhirnya, asumsi implisit dalam perumusan Campbell-Fiske bahwa sifat-sifat dan metode yang berkorelasi mungkin tidak dapat dipertahankan dalam kebanyakan situasi; lebih jauh lagi, sulit, jika bukan tidak mungkin, untuk mengevaluasi asumsi ini dengan pemeriksaan visual dari MTMM matriks (Alwin, 1974; Kalleberg & Kluegel, 1975). Analisis faktor konfirmatori dijelaskan dan diilustrasikan dalam makalah ini memungkinkan analisis yang lebih lengkap sejauh
mana masing-masing variabel dipengaruhi oleh sifat dan metode varians.