News and Articles on Health. Medical research topics.

Test-Retest Reliabilities
Structural equation models were then used to test separately whether each of the three factors hypothesized remained stable over time. The path from initial to follow-up perceived benefits of cancer screenings (pros) was significant (coefficient=.30; p<.001); suggesting that participants’ perceived screening benefits were stable over time. All items, except one (Pros_9), had significant loading weights to the “PROS” factor. Excluding Pros_9, model fit index for the remaining 8 items were satisfactory; with χ2 (99) = 312.12, RMSEA=.09 (90% CI=.08, .10), GFI=.90, IFI=.93, TLI=.91, and CFI=.93. Similarly, the path from initial to follow-up perceived cons was significant (coefficient = .75; p<.001); revealed stabled scores on perceived screening barriers. All items were loaded significantly to the “CONS” factor, except one item (Cons_7) showed negative estimates at follow-up and thus were removed. The model with the remaining 6-items fit well, with χ2 (48) = 125.81, RMSEA=.08 (90% CI=.06, .09), GFI=.92, IFI=.92, TLI=.89, and CFI=.92. Finally, the path from initial to follow-up perceived risk was also significant (coefficient = .71; p<.001), indicating participants’ perceived risk of getting cancer was stable overtime as well. The model also fits well with all items loaded significantly (χ2(6) = 11.59, RMSEA=.06 [90% CI=.00, .11], GFI=.99, IFI=.99, TLI=.97, and CFI=.99).

Confirmatory Factor Analysis (CFA)
The CFA was then applied to test the remaining 17-item three-factor model. The structure of item loadings was consistent with the intended theoretical constructs. All items measuring perceived benefits of cancer screening in general or early detection were loaded to “PROS” factor, and those measuring perceived barriers to cancer screening were loaded to “CONS” factor. In addition, items measuring perceived cancer risk were loaded to “RISK” factor. Although chi-square test was significant, the ratio of chi-square and degree of freedom was small (272 / 116=2.34), indicating good model fit (Bollen, 1989). The values of Comparative Fit Index (CFI), Incremental Fit Index (IFI), and Non-Normed Fit Index (NNFI) or Tucker-Lewis Index (TLI) were .92, .92, and .90, respectively, demonstrating adequate fit (Byrne, 1998). Furthermore, the Root Means Square Error of Approximation (RMSEA=.07; 90% CI= [.06, .08]) was small, which also indicated a good fit (Raykov, 2001).

Based on Bagozzi and Yi’ criterion (Bagozzi & Yi, 1988), all of the factor loadings, standard errors, and t ratios indicated a good fit of internal structure of model, with items of significant coefficients. The results revealed (1) no coefficients with theory contradicting signs; and (2) all standard errors seem small as indicated by large t-ratios. All t values were significantly greater than 1.96 based on Joreskog and Sorbom’s criterion (Joreskog & Sorbom, 1996). Examination of the factor coefficients revealed that all were substantially loaded by the corresponding factors. Finally, there were no negative variance estimates in the latent variable and the error covariance matrices. These results revealed no obvious mis-specifications, and supported that the hypothesized model was satisfactory. Findings supported that the CSBS-C assessed three theoretical constructs.