After a 3 week period 391 responses were received representing a 416 response

After a 3 week period 391 responses were received

• Essay
• 11

This preview shows page 4 - 6 out of 11 pages.

After a 3 week period, 391 responses were received, representing a 41.6 % response rate. Removal of responses with more than 20 % missing data resulted in 69 responses that were deemed unusable, leaving 322 usable responses for analysis. This sample size far exceeds the minimum required to provide sufficient analytical power to conduct a least squares based partial least squares model fit (Cohen 1988 ). In terms of gender, the total sample had 148 (46 %) females and 174 (54 %) males with a median age range of 21 25 years. A comparison of early and late respondents, using an independent sample t -test revealed no significant difference (Armstrong and Overton 1977 ). Fig. 1 Conceptual framework 16 K.G. Boakye et al.
4 Analyses and results To discover the underlying structure of the data, we conducted an exploratory factor analysis (EFA) (Fabrigar et al. 1999 ). Principal component analysis (PCA) with Varimax rotation (Nunnally 1978 ) was conducted to explore the underlying factors. Using the criteria of eigenvalues greater than 1 , along with examination of the scree plot and a cumulative percentage of variance explained at 79.75 %, the existence of four factors was supported (see Table 1 ). Using Harman s one-factor test (Harman 1976 ), we did not find any issue with common method bias in our dataset as none of the four exogenous variables accounted for more than half of the variance in the endogenous variable (Podsakoff et al. 2003 ). 4.1 Outer (measurement) model assessment The data was analyzed using structural equation modeling (SEM). Because our sample size was not sufficient to satisfy the power requirements for a covariance-based SEM approach and did not meet the multivariate normality assumptions associated with those methods, we used partial least squares (PLS) to analyze the data and test the overall significance of the model. PLS is a multivariate statistical technique that is widely accepted in the operations management (Peng and Lai 2012 ), information systems (Wetzels et al. 2009 ; Keil et al. 2000 ), and marketing journals (Barclay 1991 ; Hair et al. 2012 ). PLS s latent variables are weighted composite scores of the indicator variables, leading directly to explicit factor scores (Chin and Todd 1995 ). In addition, it is less restrictive on sample size and residual distribution restrictions (multivar- iate normality assumptions) than are found in other analysis models (Chin et al. 2003 ). In fact, the PLS approach is a robust method, providing both measurement and structural information in terms of indicator loadings and path coefficients. PLS parameters are estimated using a resampling approach (i.e., bootstrap or jackknife) since it lacks the classical parametric inferential framework (Peng and Lai 2012 ; Wold 1982 ). Largely, we used PLS because (1) it has the ability to test the explanatory power and predictive validity of our model, and (2) it handles formative constructs without identification problems.