Stats help! My experiment has two groups: a familiar brand group vs. an unfamiliar brand group. In the familiar brand group, I measured participants' pre-existing brand attitudes. Then I divided the familiar brand group into two: (favorable pre-existing attitude group and unfavorable pre-existing attitude group) through median split. Finally, I compared the following three groups: 1) unfamiliar brand group, 2) favorable pre-existing attitude group, and 3) unfavorable pre-existing attitude group. Reviewers say I cannot do the median split but need to treat pre-existing brand attitude as a moderator using PROCESS MACRO. However, in my view, pre-existing brand attitude is not a real moderator because I did not even measure this variable for the unfamiliar brand group (this variable is not applied to the group). Thoughts?
�������������kl���/
The way you have written your question, it is difficult to follow what your dependent variable is. But it appears your reviewers are correct. If your predictor, brand attitudes, is continuous, do not split it at the median. Keep it continuous.
As Yang Feng has pointed out, pre-existing brand attitude has not been measured for anyone in the Unfamiliar Brand group. So I think that a single analysis using the pre-existing brand attitude scores as a variable is not going to work.
What is the dependent variable? And what is the research question (in words)? Would it make sense to carry out a two-stage analysis as follows?
HTH.
Bruce Weaver, thanks for your response. I agree with you. I have already done step 1 and found no difference between Unfamiliar and Familiar brand groups (IV = brand familiarity, DV = ad attitude) when pre-existing brand attitude is not considered. I will follow step 2 and focus only on the familiar group to explore the role of pre-existing brand attitude.
However, if I still aim to compare Unfamiliar and Familiar brand groups in terms of ad attitude and I want to take into account the role of pre-existing brand attitude, what shall I do? The main argument that I am making is: Without considering pre-existing brand attitude, there is no difference between Unfamiliar and Familiar brand groups in terms of ad attitude. However, when pre-existing brand attitude is considered, there is a significant difference between Unfamiliar and Familiar brand groups in terms of ad attitude.
Because you have no pre-existing brand attitude scores for the Unfamiliar group, I can think of no standard method that allows you to partial out the effect of pre-existing attitude when comparing the Unfamiliar and Familiar groups--there is nothing to partial out for the Unfamiliar group. The only thing that comes to mind, and I am not at all convinced of its legitimacy, would be to compare raw scores for the Unfamiliar group to residuals (from the aforementioned regression analysis) for the Familiar group. Variance in the residuals from the regression analysis (performed using Familiar group only) represents variance in the DV that is not accounted for by pre-existing Brand attitudes.
HTH.
EDITED (3-Jan-201): There is a problem with what I suggested above: The mean of the residuals will always be 0. So the analysis I suggested will not work, at least not with some further tweaking. And it is not immediately obvious to me how one would tweak it.
- Badges
- Science topic
- Similar topics
- Social Psychology
- Attitude