Following the confusion (on my part) here, and as suggested, I am presenting my question in a neutral way below:

Problem: I am looking at the relationship between "product research" (variable A) and "product purchase" (variable B). 'Product' means any consumer goods (e.g. toys, clothes etc.).

A and B are composite scores (i.e. made up of responses to several questions). The compositie scores for A corresponds to a five point scale (going from no research to a lot of research), while scores for B also corresponds to a five point scale (going from decision not to buy to always buy). I need to better describe my scale labels, but they are based on a five point Likert scale.

I am treating my composite scores as interval level data. I have checked the scatterplot and there is a linear relationship between A and B.

A selection of my data set is as follows:

Variable A - Product Research: 2.98 2.98 3.12 3.14 3.16 3.32 3.19 3.21 3.32 3.34 3.46 3.55 3.72 3.76 3.96 4.03 4.14 1.21 1.65 1.78 1.72 1.08 1.04 1.98

Variable B - Product Purchase: 3.54 2.53 2.52 3.14 3.09 3.34 2.57 3.11 2.99 2.89 3.45 3.39 3.26 3.77 3.68 3.86 4.03 1.22 1.34 1.68 1.65 1.09 1.32 2.09

Question 1 How should I calculate if a relationship exists betwen product research and product purchase?

Question 2 How do I find out if certain demographic characteristics such as gender and age influence the relationship between product research and product purchase?

Note: My focus is on the overall relationship between product research and product purchase and how this relationship is shaped by demographic characteritics.

In other words, I am NOT interested in the difference between males and females on its own; rather on the difference (if any) between the relationship of A and B for males and females. (unsure if this is the same thing but to me, there is a subtle difference)

  • $\begingroup$ Presumably these data are meant to be paired in the order given. $\endgroup$
    – whuber
    Oct 28 '11 at 21:43
  • $\begingroup$ Yes, that is the intention $\endgroup$ Oct 29 '11 at 0:21

It seems clear to me that purchase must depend (in the statistical sense) on research rather than the other way round. Therefore, I'd regress that way, plus "maleness" (that is, 1 for male, 0 for female) plus the interaction between these two, plus age, plus the interaction of age and research, and so on. The coefficient on research tells you about the relationship between research and purchase, and the coefficients on the interactions tell you if this relationship varies.

I'd also plot the data. Interactions are often not easy to interpret. Both SAS and R (and probably other statistical software) makes this sort of interaction plot fairly easy.

  • $\begingroup$ Thanks. I would love a little more detail to make sense of your answer. When you say "I'd regress it that way...", what test should I use. (As suggested in the other post, I am keeping my thoughts to myself I know this sounds pretty basic but I am starting from scratch.... $\endgroup$ Oct 29 '11 at 14:03
  • 1
    $\begingroup$ When you do regression in any common statistical program, you get a bunch of standard output. This will include F test for the overall model and t-tests for the individual effects. For example, in R, the syntax would look something like m1 <- lm(purchase~research, age, sex, ageresearch, sexresearch) in SAS, you could use PROC GLM. $\endgroup$
    – Peter Flom
    Oct 29 '11 at 14:33

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.