Hello Ittezah,

Let's try to unpack your query.

0. I'd strongly urge you to consider using common factor analysis instead of PCA for identification of constructs which reflect human judgment. The PCA assumptions that all observed variance is common variance, and no specific, unique, or error variance exist are not realistic in your target domains. Consider maximum likelihood or principal axis extraction, and also consider oblique rotation (the factors are quite likely to be correlated in nature).

1. Usually, your specific research question(s) and the associated hypotheses would outline what sort of stages would be sensible for a hierarchical regression analysis. Without knowing more about your goals (as well as how you intend to quantify intention of use of cloud computing), I can't give you specific advice.

However, it is not unusual for folks to use so-called control variables (a.k.a., covariates) as the first stage of a hierarchical model, then the variables of interest (such as scores representing your chosen constructs) as variables for entry in a subsequent stage. For you, such control/covariates might include features such as industry type, job experience, and so on.

2. You might consider the use of structural equation modeling as the analytic method, as it both requires you to explicitly declare how you believe the respective manifest and latent variables inter-relate, as well as offering a wide variety of indicators to quantify how well the chosen model allows the reconstruction of the observed relationships in the sample data.

Good luck with your work.

Ittezah Ishtiaq I can see you are using TAM as a base model to study intention to use cloud computing. I will recommend using structural equation modeling (SEM)(better in SmartPLS than AMOS or MPlus) if you want to make your life easier.

Thank you for the recommendations!

Really appreciate it, could you kindly guide me as to how I should select the 'common' factor analysis option in SPSS? David Morse

Hello again Ittezah,

In spss, from the main menu bar, choose: Analyze / Dimension reduction / Factor, then in the subsequent dialog box, click the "Extraction..." button, and select the desired method using the dropdown menu bar ("Method").

David Morse

thank you for your response.

1. for the method that you just told me the following options appearing area:

principle component

unweighted least squares

generalised lease squares

Maximum likelihood

principle axis factoring

alpha factoring

image factoring

which option to select for ‘common‘ factor analysis.

moreover, for the structure equation modelling, I have computed the following equation:

Y (behavioural intention use cloud computing) = b1 (firm sizs) + b3 (industry type) + b4 (mean of perceived usefulness, ease of use and benefits) + b5 (mean of perceived risks and cost)

behavioural intention to use cloud computing is my dependent variable; firm size and industry type are my covariates and b4 (mean of perceived usefulness, ease of use and benefits) + b5 (mean of perceived risks and cost) are my independent variables. however, I am not sure as to how to show behavioural intention to use cloud computing within my model as it is included in the dataset. Any suggestions? Really appreciate your responses

Hello again Ittezah,

1. Maximum likelihood, or principal axis extraction (as mentioned in my first post).

Note that maximum likelihood is most in keeping with the computational methods used in most SEM programs (e.g., LISREL, AMOS, EQS, Calis, lavaan).

2. For your proposed model, if you're simply using direct measures (e.g., firm size) and computed scores (e.g., mean of perceived risks/costs), then your model is the same as that of a multiple linear regression model. The advantage of the SEM software is that it routinely computes a number of model-data fit measures which can be informative in evaluating the efficacy of your model.

Good luck with your work.