Boostrapping for Cronbach's Alpha and ...

[sineoriz@gmail.com]

Need Expert Guidance on Addressing Reviewer Comments for my Research Paper

Hello! SmartPLS community,

I am currently working on revisions for my research paper and have received some critical comments from reviewers that I need help addressing. Any insights or guidance you can provide would be highly appreciated. Below are the specific concerns:

1. Bootstrapping for Confidence Intervals: The reviewer has pointed out that bootstrapping should be applied not only to conduct tests for statistical significance but also to derive 95% confidence intervals for measures such as Cronbach's Alpha and CR. How can I implement this in SmartPLS?

2. Redundant Indicators in the Model: Some values in Table 2 seem to indicate redundant indicators in the model (e.g., CR values of 0.938, 0.955, and 0.935). How should I address this redundancy to improve the model's statistical properties?

3. HTMT Values and Adjusted R-values: The HTMT values suggest similarities between certain variables, and the reviewer has requested the insertion of adjusted R-values. Considering the high number of path relations, how can I ensure the accuracy of these values, particularly for behavioral intention and WOM?

4. Correction for Alpha Error Accumulation: Given many hypotheses tests in the paper, the reviewer has suggested conducting corrections like Bonferroni to manage alpha error accumulation. Is this a common practice with PLS-SEM, and how can I execute this within SmartPLS?

I recognize that these questions may be quite specific, but I hope that members of this forum with expertise in these areas might be willing to lend their assistance. Even a brief response pointing me in the right direction would be of great help.

Thank you in advance for your time and assistance.

[jmbecker]

1) You also get the confidence intervals with your bootstrapping results. If you choose "Amount of results: Complete" you will also get bootstrapping results for Cronbachs alpha etc.

2) If the reliability statistics such as Cronbachs alpha are very high, it could (but must not) imply that some of your measures have highly redundant items that are merely synonyms. It is actually more a conceptual problem of the scale and less a statistical problem that you can solve. Deleting items now will not improve the model, but it may indicate that the scale is not well developed. Such high values could also indicate poor data quality due to inattentive and low discriminatory response behavior (e.g., straightliners that have been speeding through your survey and always indicated the same response value).

3) Discriminant validity problems are hard to answer without knowing your research field, model, constructs, and data in more detailed. I think this is something that you need to solve yourself.
Adjusted R-values are provided by SmartPLS where you also find your normal r-square values.

4) It is relatively unusual to correct for alpha inflation because of complex models in PLS, because you hopefully have made your model and all your hypotheses before you collecting the data and running the model.
Correcting for alpha inflation is more common in multi-group analysis where you are making multiple comparisons or if you are adding parameters post-hoc to the model.