11.07.2015 Views

Index of Paper Presentations for the Parallel Sessions - Academy of ...

Index of Paper Presentations for the Parallel Sessions - Academy of ...

Index of Paper Presentations for the Parallel Sessions - Academy of ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

We used a 2-step structural equation modeling approach, refining <strong>the</strong> measurement model be<strong>for</strong>e analyzing <strong>the</strong> pathmodel (Anderson and Gerbing 1988). We also tested CommonMethod Variance and found no evidence. Having purified <strong>the</strong> measurement model, we first analyzed a base modelwithout <strong>the</strong> second order factor, store environment with direct paths from ambient, social and design factors to <strong>the</strong>mediators and found a poor fit. Next, we analyzed <strong>the</strong> model with <strong>the</strong> standard reflective factor structure, whichshowed a poor fit (χ2 = 860.33, df = 372, χ2/df = 2.3, RMSEA = .07, SRMR = .05, CFI = .91). While <strong>the</strong> fitimproved considerably, it was still below par. We <strong>the</strong>n analyzed <strong>the</strong> model in Figure 3, <strong>the</strong> <strong>for</strong>mative second ordermodel. We found that <strong>the</strong> fit improved fur<strong>the</strong>r (χ2 = 388.52, df= 155, χ2/df = 2.51, RMSEA = .05, SRMR = .05, CFI = .95) with all <strong>the</strong> fit-indices better than <strong>the</strong> recommendedcut-<strong>of</strong>f values (RMSEA < .06, SRMR < .08, CFI > .95). Finally, <strong>the</strong> Schmid-Leiman Factor Structure provided <strong>the</strong>best fit (as expected) compared to <strong>the</strong> models without a second order factor structure and a reflective second orderone (χ2 =664.96, df = 356, χ2/df = 1.9, RMSEA = .04, SRMR = 0.03, CFI = 0.95).DISCUSSIONIn this research, we evaluated four models, one a base model with no second order factor; two, a model with <strong>the</strong>default reflective second order factor model; three, a model with a <strong>for</strong>mative second order factor model and finally amodel with <strong>the</strong> SLS factor structure.We demonstrate <strong>the</strong> efficacy <strong>of</strong> using a <strong>for</strong>mative second order and SLS factor structure.Hence, researchers using a second order model should not jump to <strong>the</strong> conclusion that a reflective second orderfactor is <strong>the</strong> only option.We empirically demonstrate that a <strong>for</strong>mative second order model works better if <strong>the</strong>re is a conceptual basis tobelieve that <strong>the</strong> first order factors drive <strong>the</strong> second order factor.However, if some <strong>of</strong> <strong>the</strong> problems present in <strong>for</strong>mative second factors are anticipated, researchers may use <strong>the</strong> SLSfactor structure. Based on <strong>the</strong> findings in this paper, future researchers may ask <strong>the</strong> following questions:

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!