I have a logistic regression where some of the independent variables are proportions (compositions) that sum to 100 percents. Let's say that I have 4 such variables - I have ran the regressions 4 times, each time with one variable out, so that I'll be able to estimate the impact of the increase in 1% of each variable, compared to the variable that was left out. This way I can build a "ladder" of which variable is more influential than the other etc (for example - an increase of 1% in A which decreases 1% from B has OR>1 so A is more "worthy" than B). I'm now looking for an example of such method - in a paper or something similar, to see whether this is an accepted method and how to present it nicely - the Odds ratios of the 4 regressions? or a "ladder" of influence?

Any help is highly appreciated!

  • $\begingroup$ Thanks Glen, but I'm not sure what you meant. As I said - the variables that represent the percents are part of the explanatory variables - they are continuous variables that are valid to use in logistic regression.. $\endgroup$ May 1, 2014 at 4:42
  • $\begingroup$ Sorry, I missed that you were talking about explanatory variables. I've deleted the useless comment. $\endgroup$
    – Glen_b
    May 1, 2014 at 4:55


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.