Assistance with Statistical Methodology to address research problems

Im looking for some help with a statistical method(s) that can support the research questions i am looking to answer. I am basically doing research on (1) user adoption of technology and (2) whether people believe that by adopting tehnology, it can yield benefits to others (such as in education, when instructors adopt technology in their teaching activities) or in medical records keeping so that patients can chart their medical history over time.)

TO measure technology adoption I am using the UTAUT survey (Venkatesh, 2003). Unified Theory of Acceptance and Use of Technology. This is a survey with about 17 Likert questions on it (1-7) of ordinal type). The questions pertain to 4 constructs that are the determinants of user adoption (so questions 1-5 might related to a user's "performance expectancy", questions 6-9 might pertain to his "effort expectancy", and so on.

(1) The thought is that I would run Partial Least Squares regression against the data set to arrive at an equation that would set coefficients on the questions and constructs and from which I would be able to devise a measure reflecting the user's likelihood of adopting technology. SO I am applying parametic means to ordinal data but can do so since I am adding up various Likert items to define the overall constructs.

(2) Once I have the measure of user adoption, I then want to assess the relationship of that to the user's response on 1 Likert question (do you believe that by using technology, it may benefit others in some way?).

(3) Lastly, I want to see if #1 (a user's likelihood to adopt - dependent variable based on the previously discussed independent construct variables)) and #2 (the relationship between likelihood to adopt and perceived benefit upon others - independent variable) vary depending on user personality styles.

My questions are:
1) For #1, I believe I can apply a parametric test to ordinal data since I am summing ordinal data responses up, however that only applies if the data are normal. What is a good test for me to assess if the data are normal. If they are not normal, what do I do? Must I return to nonparametric tests and if so which ones might be applicable? Do I transform the data and if so what transformation method is best?

2) For #2, what statistical method would be best to use to measure the relationship between these two variables?

3) For #3, what statistical method would be best to use to assess whether different personality styles have statistically different adoption likelihoods or perceptions of whether that adoption can lead to other's benefits?

Thanks for any insights anyone can provide.


Less is more. Stay pure. Stay poor.
Reply to first question: data do not have to be normally distributed, it is their residuals. You can look at the residuals wtih a Q-Q plot and tests of normality, or if you have larger sample sizes. Yes, many people go the parametric route with combined or summed Likert style data, just check the residuals.