It's Independence Day in the U.S., so I am taking the day off, but I received the following request for advice and thought I'd pass it along to my readers.

I wonder if you could help – I am trying to create 9 different coefficient plots , which represent 9 different GLMs. In each of the 9 models, I have retained only those explanatory variables which were statistically significant. I want the plots to make clear which variables were retained and which were not.

The specifics of the situation are as follows: 9 models, 4 total covariates with 2-3 usually significant.

This leads me to throw three questions out to my readers.

1. What approach would you take to present (graphically, of course) several models that include different, but overlapping sets of covariates?
2. From your perspective, what are the benefits and costs of excluding covariates with insignificant coefficients, especially in this situation?
3. Is there an efficient way in R to skip variables in the plotting script that are not included in models? The code I've written (here and here) doesn't handle this situation well at all.

I look forward to your ideas!