# How to statistically analyze significance with 7-point Likert scale data?

I am working on a business research paper. The hypothesis is that the sales rep influences the customer's decisions. Independent variables consist of the sales rep's: knowledge, pushiness, friendliness, etc. Dependent variable is the customer's willingess to buy. In all, I have 12 questions that I asked on a Likert scale ranging from 1 (not important at all) to 7 (extremely important).

At this point, I have all the means and standard deviations calculated but am not sure how to proceed to evaluate whether the whole set of independent variables proves my hypothesis that the rep's qualities does influence the customer to buy.

I've done some reading on regression, ANOVA and chi squared but would really like to understand how I can apply them to this situation.

Additionally, I want to know if I compute the composite score of the subgroups, i.e. men versus women, how can I compare them statistically?

Thanks for any and all suggestions!

• Is the dependent variable also on a 7-item Likert scale? Commented Mar 20, 2013 at 3:28
• Hi Jason - Yes, in the sense that I first asked "how important are the following sales rep's features in your decision to buy", then followed by the respondent answering the following questions based on 1-7. Is that what you mean? Commented Mar 20, 2013 at 3:45
• Just to clarify: Do all data come from the customer? They rate the sales rep's knowledge, etc., then you ask them if they are willing to buy? Commented Mar 20, 2013 at 3:49
• Hi Jason - Yes they do all come from the customer. However I indirectly ask if they are willing to buy based on how they rate the questions. So the summation of the answers would indicate they are feel strongly to buy. Commented Mar 20, 2013 at 4:00
• There's a problem with the underlying hypothesis: if the sales rep is really great, the customer will never notice it was not her own decision. Non-response rates are going to be significant, as well. Multinomial logit is (superficially) what you are probably after. Please be aware there's a lot more for you to read to avoid the hidden reefs... Commented Mar 20, 2013 at 6:56

As you have already identified dependent and independent variables, you can use correlation and regression analysis. If you have qualitative variables you can also conducted comparison tests such as t-test for two groups and ANOVA for more than two group qualitative variable.

You can group your data and perform either a 2-proportion test or a Chi-square test.

Ho: Sales Rep knowledge has no impact on customer buying criteria Ha: Knowledgable Sales Rep has a positive impact on customer buying criteria

A simple grouping:

Divide your data into 2 groups (low sales rep knowledge and high sales rep knowledge). Divide your results into 2 groups (will likely buy and will not buy).

Once you have this grouped data then you can perform a 2-proportion test.

If you have more than 2 groups, then a Chi-square test would be appropriate.

• Dividing your individuals into two groups is a bad idea as it discards information and reduces power. Commented Dec 13, 2013 at 20:05