I have carried out MR with three continuous predictors of which only one is sig. Is it OK to illustrate this visually in a partial plot as produced by SPSS ie in a plot of A (outome) x B (sig predictor) while accounting for C and D (the two ns predictors)? I could do a standard AxB scatterplot but the partial plot seems to me to be a truer expression of the effect. I ask because I’m presenting to a mixed audience of people who understand regression or are non statisticians and I want to satisfy the former but still be able to communicate the effect to the latter. Thank you.
I think that people would appreciate the standard AxB plot more than a partial plot. It is more intuitive and easier to interpret even for statisticians. I’ve found that presenting material that requires people to think at two or more levels in order to understand what I am presenting is testing fate. Once I mentioned to a group of docs that my test of homoschedasticity failed and so I compensated by running a certain type of test. Attempting to explain homoschedasticity threw them off. All they wanted to know was were the results significant or not. Oh well.
I would strongly advise you to work with ‘effect’ sizes (substantive values like R-squared) in deciding which predictors to retain or display. ‘Significance’ is misleading (or just plain wrong) and most audiences will quite correctly not understand what you are talking about.