In any form of regression model, we often think of the effects as additive. That is, we suppose that the effect of one variable can be added to the effect of another to get an accurate model. This is never strictly true, but how true is it? Is it true enough? How can we tell? Read more!
One question that sometimes arises in doing statistical analysis is whether to use a sophisticated method that is (in one way or another) more appropriate than a more typical method. The reason for its appropriateness might be that the usual method violates assumptions (e.g. we should use robust regression rather than OLS regression in some cases), answers the question better (e.g. we might use quantile regression instead of OLS regression in some cases), or is more efficient.
But the reviewers and editors at a journal may not know of the new method and may have issues with it. It might even lead to the paper being rejected.
What are your thoughts on this?
You’re about to do some research. You’ve got an idea in your field and you hope to turn it into a grant or an article or a dissertation. Right now, you may not know exactly how you want to work with the idea. It’s just a thought.
How can a statistician help you?
At this early stage, the consultant may be able to suggest novel ways of testing your idea, using analytic methods you’ve never heard of. Then, the statistical consultant will be able to help you plan a sample size and a sampling strategy, write up a methods section and start gathering and recording data in a sensible way.
After you gather data, your statistical consultant will be able to help you analyze the data and understand it, using methods that are appropriate rather than just some method you learned in your two statistics courses.
For instance, if you have a dependent variable and one or more independent variables you may want some sort of regression. But is it
But maybe you don’t need regression at all; or maybe you need a multilevel model. Perhaps factor analysis or principal component analysis or cluster analysis or multidimensional scaling will be vital to answering your research question.
Your statistical consultant will be able to help you figure it out.
Two terms that are frequently confused are moderation and mediation:
Definitions of moderating and mediating:
A mediating variable is one that accounts for or alters another relationship. A strict definition is that the mediating variable has to greatly reduce or even eliminate the relationship. A more lenient definition is that it affects the relationship. I favor that more lenient definition. A moderating variable is one that interacts with a variable. An interaction means that the relationship between one independent variable (IV) and the dependent variable (DV) is different at different levels of the other IV. Read more!
In regression and ANOVA, an interaction occurs when the effect of one independent variable on the dependent variable is different at different levels of another independent variable. When one or both of the independent variables is categorical, then two common strategies for dealing with interactions are stratifying and adding an interaction term. A somewhat less common method is classification and regression trees. Each has its advantages and disadvantages.